Systems and methods for label generation in online learning of channel estimation for demodulation

Abstract

A wireless transmit / receive unit (WTRU) may receive channel estimation training allocation information. The channel estimation training allocation information may include one or more of L-DMRS RE allocation information and / or L-DMRS power offset information. The WTRU may receive L-DMRS REs and DMRS REs, for example according to the channel estimation training allocation information. The WTRU may determine whether online training is complete, for example based on the L-DMRS REs and / or the DMRS REs. The WTRU may send, for example based on the determination of whether online training is complete, an indication of whether online training is complete to a network.

Description

SYSTEMS AND METHODS FOR LABEL GENERATION IN ONLINE LEARNING OF CHANNEL ESTIMATION FOR DEMODULATIONCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of United States Provisional Application No. 63 / 610,771 filed on December 15, 2023, the entire contents of which are incorporated herein by reference.BACKGROUND

[0002] Artificial intelligence (Al) may include 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, for example without explicit configuration of sequence of steps of actions. An Al component may enable learning complex behaviors which, for example may be difficult to specify and / or implement when using legacy methods.SUMMARY

[0003] A wireless transmit / receive unit (WTRU) may receive a learning demodulation reference signal (L-DMRS) resource and / or a L-DMRS configuration. The WTRU may receive a DM RS resource (RE) and / or a DMRS configuration. The WTRU may input the L-DMRS and / or the DM RS resource to a high-quality-input-based channel estimator. Based on the L-DMRS resource and / or the DMRS resource for example, the WTRU may generate a label from the high-quality-input-based channel estimator. The label may be for an artificial intelligence / machine learning (AI / ML) model. The WTRU may use the label to train an AI / ML model. Based on the label for example, the WTRU may train the AI / ML model. The WTRU may transmit a training complete message to a network device.

[0004] The WTRU may additionally, or alternatively, receive data resources and / or input the data resources to the AI / ML model. The WTRU may generate a prediction from the high-quality- input-based channel estimator. The prediction may include one or more of a channel estimate at the DMRS resource, a channel estimate at the L-DMRS resource, a noise variance, and / or a Doppler estimate. The WTRU may measure a rate of change of a loss function and / or compare the rate of change to a threshold value. For example, transmitting the training complete message to a network device may be based on measurement of the rate of change of the loss function and / or a comparison of the rate of change to the threshold value.

[0005] A wireless transmit / receive unit (WTRU) may receive channel estimation training allocation information. The channel estimation training allocation information may include one or more of L-DMRS RE allocation information and / or L-DMRS power offset information. The WTRU may receive L-DMRS REs and DMRS REs, for example according to the channel estimation training allocation information. The WTRU may determine whether online training is complete, for example based on the L-DMRS REs and / or the DMRS REs. The WTRU may send an indication of whether online training is complete (e.g., to a network), for example based on the determination of whether online training is complete. The indication may include a request for retraining, for example if the WTRU determines that online training is not complete. The indication may include an indication that online training is complete, for example if the WTRU determines that online training is complete.

[0006] The WTRU may determine whether online training is complete based on an error associated with the L-DMRS REs and / or DMRS REs. Additionally, or alternatively, the WTRU may determine whether online training is complete based on a rate of change of a loss function being less than a threshold. For example, if an error and / or rate of change of a loss function is greater than or equal to a threshold, the WTRU may determine that online training is not complete. If, for example, the error and / or rate of change of the loss function is less than the threshold, the WTRU may determine that online training is complete. In some examples, the WTRU may determine that online training is complete if the error and / or rate of change of the loss function is less than a first threshold and a second threshold.

[0007] The WTRU may generate high-quality labels, for example based on the pre-processing of the L-DMRS REs and DMRS REs. The WTRU may train an artificial intelligence / machine learning (AI / ML)-based channel estimator, for example based on the generated high-quality labels. In some examples, the WTRU may send a training complete message to a network, for example based on a determination that the AI / ML-based channel estimator is sufficiently trained.

[0008] The WTRU may determine to initiate online learning of channel estimation for demodulation, for example based on a location change. In some examples, the WTRU may send a request for the channel estimation training allocation information. The WTRU may train the AI / ML-based channel estimator based on a difference between the generated high-quality labels and predictions from the AI / ML-based channel estimator. In some examples, the WTRU may determine that the AI / ML-based channel estimator is sufficiently trained based on a rate of change of a loss function being less than a threshold.

[0009] The WTRU may send a request for retraining of the AI / ML channel estimator, for example to a network. The WTRU may determine to send the request for retraining of the AI / ML channel estimator based on an indication from the network requesting retraining. Additionally, or alternatively, the WTRU may determine to send the request for retraining of the AI / ML channel estimator based on a difference between the generated high-quality labels and predictions from the AI / ML-based channel estimator being less than a threshold.BRIEF DESCRIPTION OF THE DRAWINGS

[0010] FIG. 1A is a system diagram illustrating an example communications system in which one or more disclosed embodiments may be implemented.

[0011] 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.

[0012] 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. 1A according to an embodiment.

[0013] 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. 1A according to an embodiment.

[0014] FIG. 2 is an example NR DMRS structure with configuration type 1 for ports 1000-1003.

[0015] FIG. 3 is an example NR DMRS structure with configuration type 2 for ports 1000-1003.

[0016] FIG. 4 is an example of DMRS-based channel estimation, equalization, demodulation, and channel decoding.

[0017] FIG. 5 is an example of ReEsNet model architecture.

[0018] FIG. 6 shows an example of a system with a WTRU receiving the normal density DMRS REs and / or low density DMRS REs.

[0019] FIG. 7 is an example of resource allocation for channel estimation training and / or performance monitoring.

[0020] FIG. 8 is an example of resource allocations during AI / ML model inference.

[0021] FIG. 9 is an example port structure of L-DMRS and DMRS (e.g., configuration type 1) with 4 layers.

[0022] FIG. 10 is an example port structure of L-DMRS and DMRS (e.g., configuration type 2) with 12 layers.

[0023] FIG. 11 is an example port structure of L-DMRS and low density DMRS with 1 layer.

[0024] FIG. 12 is an example port structure of L-DMRS and low density DMRS with 8 layers.

[0025] FIG. 13 is an example port structure of DMRS and variable L-DMRS and data pattern with 1 layer.

[0026] FIG. 14 is an example of AI / ML model training.

[0027] FIG. 15 shows an example procedure for the WTRU to monitor the performance of the AI / ML-based channel estimation.

[0028] FIG. 16 is an example OFDM resource grid with L-DMRS and DMRS allocations.

[0029] FIG. 17 is an example channel estimation normalized mean square error (NMSE) for AI / ML-based and non-AI / ML-based channel estimators.

[0030] FIG. 18 is an example of a procedure for label generation, training, and performance monitoring of uplink channel estimation.DETAILED DESCRIPTION

[0031] 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 access system 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.

[0032] As shown in FIG. 1A, 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 “STA”, may be configured to transmit and / or receive wirelesssignals 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.

[0033] 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 Internet 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 be appreciated that the base stations 114a, 114b may include any number of interconnected base stations and / or network elements.

[0034] 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.

[0035] 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).

[0036] 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).

[0037] In an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement a radio technology such as Evolved UMTS Terrestrial Radio Access (E-UTRA), 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).

[0038] In 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).

[0039] 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).

[0040] 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.

[0041] 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 cellularbased 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.

[0042] 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.

[0043] 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 ownedand / 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.

[0044] 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.

[0045] 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.

[0046] 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 Specific Integrated 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. 1 B 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.

[0047] 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 totransmit 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.

[0048] 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.

[0049] 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.

[0050] 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 may include 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).

[0051] 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.

[0052] 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 location-determination method while remaining consistent with an embodiment.

[0053] 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.

[0054] 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 an interference 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)).

[0055] 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.

[0056] 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 oneembodiment, 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.

[0057] 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.

[0058] The CN 106 shown in FIG. 1C 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.

[0059] 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.

[0060] 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 WTRUs 102a, 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.

[0061] 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.

[0062] 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 providethe 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.

[0063] Although the WTRU is described in FIGS. 1A-1D 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.

[0064] In representative embodiments, the other network 112 may be a WLAN.

[0065] 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 802.11e DLS or an 802.11z tunneled DLS (TDLS). A WLAN using an Independent BSS (IBSS) mode may not have an AP, and the STAs (e.g., all of the STAs) within or using the IBSS may communicate directly with each other. The IBSS mode of communication may sometimes be referred to herein as an “ad-hoc” mode of communication.

[0066] 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 STA), including the AP, may sense the primary channel. If the primary channel is sensed / detected and / or determined to be busy by a particular STA, the particular STA may back off. One STA (e.g., only one station) may transmit at any given time in a given BSS.

[0067] 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.

[0068] 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).

[0069] 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 802.11af and 802.11 ah relative to those used in 802.11 n, and 802.11ac. 802.11af supports 5 MHz, 10 MHz and 20 MHz bandwidths in the TV White Space (TVWS) spectrum, and 802.11ah supports 1 MHz, 2 MHz, 4 MHz, 8 MHz, and 16 MHz bandwidths using non- TVWS spectrum. According to a representative embodiment, 802.11ah 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 / or limited bandwidths. The MTC devices may include a battery with a battery life above a threshold (e.g., to maintain a very long battery life).

[0070] WLAN systems, which may support multiple channels, and channel bandwidths, such as 802.11 n, 802.11ac, 802.11af, 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 802.11ah, 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.

[0071] 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.

[0072] 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.

[0073] 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 carriers may 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).

[0074] 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).

[0075] 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.

[0076] 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. 1 D, the gNBs 180a, 180b, 180c may communicate with one another over an Xn interface.

[0077] The CN 115 shown in FIG. 1D 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.

[0078] 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 (notshown) that employ other radio technologies, such as LTE, LTE-A, LTE-A Pro, and / or non- 3GPP access technologies such as WiFi.

[0079] The SMF 183a, 183b may be connected to an AM F 182a, 182b in the ON 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.

[0080] 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.

[0081] 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 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. 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.

[0082] In view of Figures 1A-1 D, and the corresponding description of Figures 1A-1D, 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.

[0083] 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 oneor 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.

[0084] 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.

[0085] Systems and methods for the WTRU to enable online learning of the channel estimation for demodulation as described herein, for example for systems using AI / ML models in channel estimation for demodulation. Additionally, or alternatively, AI / ML models may be used, for example for improved system performance and / or reduced DM RS overhead size. A channel estimator may be utilized / learned. For example, the channel estimator may be utilized / learned offline. The training may be (e.g., generally) performed with the error-free ground truth dataset, for example when the channel estimator is utilized / learned offline. Offline learning with an offline dataset may, for example, not adapt to the environment and / or the radio impairment(s).

[0086] A WTRU may be configured to use an AI / ML model in channel estimation for demodulation. For example, the WTRU may request online training of downlink channel estimation for demodulation. The WTRU may be configured to use (e.g., new) learning demodulation reference signals (L-DMRS). The L-DMRS may be orthogonal to known pilot sequences, for example with the same orthogonality as the DM RS. The L-DMRS may have higher power and / or higher density other DMRS (e.g., as in 5G NR). The L-DMRS may be utilized to create the high-quality labels for the effective channel estimates, which for example may be used during training of the AI / ML-based channel estimator. The WTRU may be configured to use L-DMRS in the performance monitoring of the AI / ML-based channel estimator for demodulation. The WTRU may be configured to transmit the L-DMRS for the labelgeneration for online learning and / or performance monitoring of uplink channel estimation for demodulation.

[0087] Artificial intelligence (Al) may be configured for behavior exhibited by machines that, for example may mimic cognitive functions to sense, reason, adapt, and / or act. An Al component may include the realization of behaviors and / or conformance to requirements by learning based on data, for example without explicit configuration of sequence of steps of actions. An Al component may enable learning complex behaviors. Complex behaviors may be difficult to specify and / or implement, for example when using legacy methods.

[0088] Machine learning (ML) may include algorithms (e.g., types of algorithms) that are configured to solve a problem based on learning through experience (e.g., data), for example without being explicitly programmed (e.g., configuring a set of rules). ML may be a subset of Al. Based on the nature of data or feedback available to the learning algorithm for example, different ML paradigms may be envisioned. A supervised learning approach may involve learning a function that maps input to an output based on labeled training example.

[0089] In some examples a (e.g., each) training example may include a pair. The pair may include an input and a corresponding output. An unsupervised learning approach may involve detecting patterns in the data with no pre-existing labels. A reinforcement learning approach may involve performing sequence of actions in an environment, for example to maximize the cumulative reward. ML algorithms may be applied using a combination and / or interpolation of the approaches herein. 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. Semi-supervised learning may fall between unsupervised learning (e.g., with no labeled training data) and supervised learning (e.g., with only labeled training data).

[0090] Deep learning may include a class of ML algorithms that may employ artificial neural networks, for example Deep Neural Networks (DNNs). DNNs may be (e.g., loosely) inspired by biological systems. The DNNs may be a special class of ML models that may be inspired by the human brain. The input may be linearly transformed and / or pass through non-linear activation function multiple times. DNNs may include multiple layers. A (e.g., each) layer may include a linear transformation and / or a given non-linear activation function. The DNNs may be trained using the training data, for example via back-propagation algorithm. DNNs may be utilized in a variety of domains including, for example speech, vision, natural language, wireless communication, and / or etc., and / or for (e.g., various) ML settings, for example supervised, unsupervised, semi-supervised, and / or etc.

[0091] A demodulation reference signal (DM RS) may be utilized. Coherent demodulation of signals transmitted over the radio interface may utilize knowledge of an effective (e.g., precoded) wireless channel. The channel estimation process at the receiver in NR may utilize the transmission of physical channels accompanied with demodulation reference signals (DMRS). DMRSs may be generated using pseudo-random sequences, for example based on systems parameters known to the receiver. The parameters used to control the sequence generation may include scrambling identity, symbol locations, number of OFDM symbols in a slot, and / or etc. The DMRS operation in NR may include (e.g., several) predefined options for patterns (e.g., uniform / equally spaced) and / or densities of RSs, for example based on one or more of the physical channels. Additionally, or alternatively, the DMRS operation may be configured using scheduling (e.g., DCI-based) and / or high-layer configuration, for example for different use cases and / or WTRU capabilities.

[0092] The configuration of the DMRS may include one or more of a density and / or pattern in the resource grid, a duration, a starting symbol (e.g., front-loaded DMRS), and / or cover code(s), for example to differentiate between antenna ports sharing the same time / frequency resources (e.g., for single-user and multi-user MIMO cases). The set of parameters for DMRS may be different, for example depending on the physical channel and / or depending on WTRU capability. For example, for PDSCH DMRS, there may be one or more of configuration type 1 or type 2, mapping type A or type B, starting symbol for mapping type A, single versus double symbol DMRS, DMRS additional positions, and / or duration.

[0093] The (e.g., specific) selection of DMRS may be carried out by (e.g., both) higher-layer configuration and / or dynamic (e.g., DCI-based) signaling. Additionally, or alternatively, there may be a default configuration. The gNB may signal (e.g., using RRC, MAC-CE, and / or PDCCH / DCI) the selection to the terminal, for example upon selection of DMRS settings. FIGs. 2 and 3 show example DMRS patterns 200, 300 over one slot and one resource block in NR with CDM grouping across the frequency and code domains. In FIG. 2, a DMRS pattern 200 follows mapping type A, configuration type 1 , and starting symbol 3 using downlink antenna ports 1000-1003. In FIG. 3, a DMRS pattern 300 uses mapping type A, configuration type 2 and starting symbol 2 using downlink antenna ports 1000-1003.

[0094] The terminal may (e.g., then) utilize the DMRS for channel estimation and / or coherent demodulation of the corresponding physical channels. The terminal may utilize the DMRS for channel estimation and / or coherent demodulation of the corresponding physical channels through specific receiver filter implementation (e.g., least squares, minimum mean squared error (MMSE), and / or etc.) which, for example may (e.g., broadly) estimate the composite channel.Estimating the composite channel may include mapping the transmitted layers onto the receive antennas, for example for the resource blocks that are scheduled.

[0095] FIG. 4 shows an example channel estimation process 400. At 402 the receiver may (e.g., first) determine the channel estimate(s) of the DMRS symbol(s), for example from their known locations in the received slots. The receive may use an averaging window to minimize the effects of noise. At 404 the receiver may (e.g., then) use multi-dimensional interpolation operations, for example to estimate the missing values associated with (e.g., all) the other resources (REs) from the channel estimation grid. At 406 the receiver may estimate noise power, for example to improve performance by comparison of direct and average channel estimates. Based on the channel and / or noise estimates for example, at 408 the terminal may (e.g., then) design an equalizer (e.g., MMSE). At 410, the terminal may (e.g., then) perform coherent demodulation and / or channel decoding, for example based on channel and / or noise estimates. At 414 the receiver may output information bits and / or the procedure may end.

[0096] The performance of the receiver functions (e.g., equalization, demodulation, channel decoding, and / or etc.) may be (e.g., directly) influenced by the quality of the effective and / or precoded channel estimate. The receiver may utilize a (e.g., significant) number of DMRS REs, for example for achieving a desired (e.g., satisfactory) performance in the receiver functions. A high DMRS overhead may reduce spectral efficiency. The receiver may decrease DMRS overhead, for example by (e.g., merely) reducing the number of DMRS REs in one or more (e.g., each) RB. System performance may suffer due to poorer channel estimation performance. The DMRS signals across various layers / users may be orthogonal for multi-layer transmissions (e.g., SU-MIMO and / or MU-MIMO), which for example may (e.g., further) increase the DMRS overhead. For example DMRS overhead may increase as the number of layers and / or coscheduled users increases. Conventional channel estimation techniques may be subject to high DMRS overhead for adequate system performance, which for example may lead to lower spectral efficiency. Allocating a lower number of DMRS symbols may negatively affect the system performance, for example due to the poor channel estimates obtained via the conventional channel estimation techniques. AI / ML may be used to learn the channel estimation for demodulation, for example for improved system performance and / or reduced DMRS overhead size. Training may be performed with the error-free ground truth channel dataset, for example when the channel estimation functions are learned offline. The offline training dataset may be obtained from (e.g., a handful of) available / agreed channel models, for example (e.g., then) the dataset will be limited to a set of environments without including the radio impairments. The offline training dataset may not be well matched to the local environment where a deviceresides in some examples. The offline learning of channel estimator for demodulation with the offline dataset may not adapt to the environment as well as the radio impairments.

[0097] In the downlink for example, the WTRU may request online training of downlink channel estimation for demodulation. The WTRU may be configured to use one or more new learning demodulation reference signals (L-DMRS). The L-DMRS may be orthogonal known pilot sequences with the same or similar orthogonality as the DMRS. The L-DMRS may have higher power (e.g., level) and / or higher density in the time and / or frequency domain compared to the DMRS, for example as in 5G NR. The WTRU may utilize L-DMRS to create the high-quality labels for the effective channel estimates, which for example may be used during training of the AI / ML-based channel estimator. The WTRU may be configured to use high-quality labels in the performance monitoring of the AI / ML-based channel estimator for demodulation. Additionally, or alternatively, the WTRU may be configured to transmit the L-DMRS for the label generation for online learning and / or performance monitoring of uplink channel estimation for demodulation.

[0098] Systems and methods for online learning of DL channel estimation for demodulation are disclosed herein. The WTRU and / or the network (NW) AI / ML endpoint (e.g., a gNB, a cloud server, and / or an edge computing server) may detect and / or be informed of performance degradation of an AI / ML model. When a WTRU (e.g., first) enters a new geographic region, the AI / ML model may not be well trained for the (e.g., new) local environment. The WTRU may request support for retraining / updating the AI / ML model, for example for the new environment. Moving to the local environment / geographic region may include an environment and / or location change. The WTRU may retrain the AI / ML model online, for example with the assistance of another node in the network (e.g., gNB, another WTRU using sidelink or WTRU-to-WTRU direct communication, and / or etc.). Online learning of an AI / ML-based downlink channel estimator for demodulation may (e.g., then) enable true adaptation to the local environment and / or the radio impairments. Labels for an AI / ML-based downlink channel estimator may be created via new learning DMRS (L-DMRS) pilots and / or NR PDSCH DMRS pilots. The NR PDSCH DMRS pilots may be referred to as DMRS pilots herein. The WTRU may perform performance monitoring of the AI / ML model, for example via the L-DMRS. A WTRU may generate the labels for the online learning of the AI / ML-based downlink channel estimator and / or perform performance monitoring, for example via L-DMRS.

[0099] Online learning and label generation may be performed. The WTRU and / or the NW may detect a need for initiating the online learning of channel estimation for demodulation (e.g., by systems and / or methods described herein). The WTRU may request online training of channel estimation for demodulation. The request may include capabilities for use of different types ofeffective channel estimation training allocations (e.g., type1 / type2), and / or if a data RE may be used as input to the AI / ML-based channel estimator. The NW may respond, for example by one or more of increasing the available non-specific effective channel estimation training allocations, indicating to the WTRU non-specific effective channel estimation training allocation information, and / or scheduling WTRU specific effective channel estimation training allocations.

[0100] The WTRU may be configured to use one or more learning demodulation reference signals (L-DMRS). The L-DMRS may include orthogonal pilot sequences, for example with a same orthogonality as the DMRS. The L-DMRS pilots may have higher power (e.g., level) and / or higher density in the time and / or frequency domain compared to the DMRS pilots (e.g., as in 5G NR). The WTRU utilizes L-DMRS to create the high-quality labels for the effective channel estimates. The WTRU uses these labels during training of the trainable AI / ML-based channel estimator.

[0101] FIG. 5 shows an example ReEsNet model architecture 500, which for example may be constructed via one or more of residual blocks (ResBlock), convolutional (conv) layers, rectified linear unit (ReLU) activation layers, and / or residual connections. For example, a ResBlock may include one or more convolutional layers and / or one or more ReLU activation layers. Other variations of ReEsNet architecture may be utilized.

[0102] At 502 one or more DMRS REs may be received. At 504 the received one or more DMRS REs may be preprocessed. At 506 the preprocessing may be complete and / or the preprocessed DMRS REs may be output.

[0103] The WTRU may preprocess by extraction of the received DMRS REs from the received signals using the configured DMRS locations (e.g., allocations). Additionally, or alternatively, the WTRU may preprocess by extraction of the received data REs from the received signals using the configured data locations (e.g., allocations). In some examples, the WTRU may preprocess by one or more of division of the received DMRS REs by known DMRS pilots, multiplication of the received DMRS REs by conjugate of the known DMRS pilots, application of a 2D filtering to the DMRS REs and / or data REs, saving DMRS REs and / or data REs in a replay buffer, resizing the input data shape, and / or concatenation of the real and / or imaginary parts to obtain the real- valued input to the AI / ML model.

[0104] At 508 the (e.g., preprocessed) DMRS REs may be input into an AI / ML model. The output of the AI / ML model may be postprocessed at 510. Postprocessing may include one or more of conversion of the real-valued AI / ML model output to the complex-valued output data and / or resizing the AI / ML model output. In some examples, postprocessing at 510 may be omitted. Output from the post-processing at 510 may include effective channel estimates at a(e.g., each) RE associated to the received DMRS REs at the AI / ML model input, for example at 512. Training may be performed offline in a supervised manner. The features may include the received DMRS REs. The labels may include the error-free ground-truth effective channel estimates at each RE.

[0105] New learning demodulation reference signal (L-DMRS) may be utilized for the online learning and the performance monitoring of the AI / ML model, for example when an AI / ML model is utilized at the WTRU for the downlink channel estimation for demodulation and / or at the NW for the uplink channel estimation for demodulation. The properties of new L-DMRS may include orthogonal known pilot sequences with the same and / or similar orthogonality as the conventional DMRS pilots. Additionally, or alternatively, properties of the L-DMRS may include higher power (e.g., level) and / or higher density in the time and / or frequency domain, for example compared to the conventional DMRS REs. Additionally, or alternatively, the L-DMRS and / or the conventional DMRS may be (e.g., jointly) utilized to generate the high-quality labels. The high-quality labels may include one or more of the effective and / or precoded channel estimates at each RE, the effective noise variance, and / or the Doppler estimate. The high- quality labels generated via L-DMRS may be used, for example during the online training and / or performance monitoring of the trainable AI / ML-based channel estimator.

[0106] FIG. 6 shows an example of a system 600 with a WTRU receiving the normal density DMRS REs (e.g., type 1a or type 2a) and / or low density DMRS REs (e.g., type 1 b or type 2b). Additionally, or alternatively, the WTRU may apply preprocessing. In FIG. 6, the WTRU may receive the L-DMRS REs and / or apply (e.g., some) preprocessing. The WTRU may receive the data REs (e.g., only for type 2a and / or type 2b training allocation) and / or apply (e.g., some) preprocessing. The WTRU may use the received L-DMRS REs and / or the DMRS REs at the high-quality-input-based channel estimator, for example during the AI / ML model training and / or performance monitoring as demonstrated in FIG. 6.

[0107] At 602 a gNB may send one or more L-DMRS REs, DMRS REs, and / or data to the WTRU, for example for preprocessing. The data may include downlink data for the WTRU. Additionally, or alternatively, the gNB may send data to the WTRU, for example for preprocessing. At 604 the WTRU may train an AI / ML model for channel estimation. For example, the WTRU may preprocess the one or more L-DMRS REs, DMRS REs, and / or data. In some examples, preprocessing may be omitted. The L-DMRS REs may be input into a high- quality-input-based channel estimator, for example after preprocessing. Additionally, or alternatively, one or more of an L-DMRS pattern and / or DMRS pattern may be preprocessed and / or input into the high-quality-input-based channel estimator. The high-quality-input-basedchannel estimator may generate labels, for example from one or more of the L-DMRS REs, DMRS REs, the L-DMRS pattern and / or the DMRS pattern. The labels may include one or more of effective channel estimates (e.g., at each RE), noise variance, and / or Doppler estimate. The WTRU and / or network may monitor the training progress, for example by measuring the rate of change of the loss function and / or comparing to a threshold. A loss function may be used as input to the high-quality-input-based channel estimator, for example during training.

[0108] Additionally, or alternatively, at 604 the WTRU may input DMRS REs and / or data into an AI / ML-based channel estimator, for example after preprocessing. Additionally, or alternatively, one or more of an L-DMRS pattern and / or DMRS pattern may be preprocessed and / or input into the AI / ML-based channel estimator. The AI / ML-based channel estimator may generate predictions, for example from one or more of the DMRS REs, the data, the L-DMRS pattern and / or the DMRS pattern. The predictions may include one or more of effective channel estimates (e.g., at each RE), noise variance, and / or Doppler estimate. The WTRU and / or network may monitor the training progress, for example by measuring the rate of change of the loss function and / or comparing to a threshold. A loss function may be used as input to the AI / ML-based channel estimator, for example during training.

[0109] The WTRU may apply preprocessing to the received L-DMRS REs and / or the DMRS REs, for example before forwarding them to the high-quality-input-based channel estimator. The high-quality-input-based channel estimator may employ high-quality and / or quantity of input RSs and / or generate the labels. The high-quality-input-based channel estimator may include 2D filtering (e.g., up-sampling, denoising, and / or etc.) and / or interpolation. For example, in type 2 (e.g., FIG. 7), the high-quality-input-based channel estimator may use only the L-DMRS REs and / or the DMRS REs as input (e.g., data REs may not be used by the high-quality-input-based channel estimator). An up-sampling filter may be used at the output, for example to provide the effective channel estimate in (e.g., all) REs. The labels may include one or more of the effective and / or precoded channel estimates at a (e.g., each) RE, an effective noise variance, and / or a Doppler estimate. The labels may be used for training the AI / ML-based channel estimator.

[0110] The WTRU may use the received DMRS REs as input to the trainable channel estimator for demodulation (e.g., AI / ML-based channel estimator), for example as in FIG. 6. FIG. 7 is an example of resource allocation for channel estimation training and / or performance monitoring. The WTRU may additionally or alternatively use the received data REs as input to the trainable channel estimator for demodulation, for example (e.g., only) for the type 2 training allocation (e.g., as in FIG. 7). During inference for example, the same subset of data REs may additionally, or alternatively be used as input. The WTRU may apply preprocessing to thereceived DMRS REs and / or the data REs, for example before forwarding them to the AI / ML- based channel estimator. The predictions and / or outputs of an AI / ML-based channel estimator may include one or more of the effective and / or precoded channel estimates at each RE, the effective noise variance, and / or the Doppler estimate.

[0111] The WTRU may perform different preprocessing to the received DMRS REs for the high- quality-input-based channel estimator and / or AI / ML-based channel estimator. The WTRU may train the AI / ML-based channel estimator, for example based on the generated labels (e.g., effective channel estimates, effective noise variance, and / or Doppler estimate). Example predictions are shown in FIG. 6. The WTRU may monitor the training progress. For example, the WTRU may measure the rate of change of the loss function and / or compare to a threshold. The WTRU may send a training complete message to the gNB, for example when the WTRU detects that training is sufficient.

[0112] FIG. 8 is an example 800 of resource allocations during AI / ML model inference. The gNB may stop effective channel estimation training allocations (e.g., no allocation for L-DMRS REs as shown in FIG. 8), for example upon receiving the training complete message. The WTRU may employ the trained AI / ML-based channel estimator for online inference. For the online inference for example, there may be no allocations for L-DMRS REs (e.g., as shown in FIG. 8). The input of AI / ML model may be the received DMRS REs and / or data REs. The output of AI / ML model may include predictions of one or more of the effective channel estimates at a (e.g., each) RE, the effective noise variance, and / or the Doppler estimate.

[0113] The WTRU may be configured to use non-specific effective channel estimation training allocations. In some examples non-specific effective channel estimation training allocations may be non-specific to any WTRU and / or multiple WTRUs may simultaneously use the same training allocations. Type 1 (e.g., type 1a or type 1b) may include non-specific effective channel estimation training allocations, for example including L-DMRS as in FIG. 7. Type 2 (e.g., type 2a or type 2b) may include non-specific effective channel estimation training allocations, for example including L-DMRS and / or data as shown in FIG. 7. In type 1a and / or type 2a for example, the allocated RBs and / or RBGs may carry normal density DMRS REs. In type 1b and / or type 2b for example, the allocated RBs and / or RBGs may carry low density DMRS REs.

[0114] The WTRU may receive information, for example from a network, to learn the current training allocation schedules. The WTRU may read the information (e.g., SIB) or RRC and / or may use the received information to learn the semi-static schedule of effective channel estimation training allocations (e.g., type 1a, type 1b, type 2a, and / or type 2b). The WTRU may be assigned to an L-DMRS group, for example by the network. For example, a WTRU may beassigned to a particular semi-static schedule of effective channel estimation training allocations by DCI.

[0115] Systems and methods for performance monitoring via L-DMRS are disclosed herein. The WTRU may request and / or be configured to monitor the performance of the AI / ML-based channel estimator, for example for demodulation. The WTRU may be configured to use L- DMRS for performance monitoring, as in FIG. 7 for example. The WTRU may be configured to use type 1a, type 1b, type 2a, and / or type 2b effective channel estimation performance monitoring allocations. The monitoring request may indicate one or more of a test bandwidth (BW) or number of resource blocks (RBs), the type of effective channel estimation performance monitoring allocations (e.g., type 1a, type 1 b, type 2a, or type 2b), and / or a performance threshold. The network (NW) may use the request to determine how large the monitoring allocations should be, for example for the test BW and / or number of RBs. The WTRU may receive a performance threshold, for example for the estimated normalized mean square error (NMSE) of the channel estimation. The WTRU may receive the performance threshold from the network (e.g., gNB), for example via SIB or RRC signaling. The WTRU may continue to use the same AI / ML model, for example if an estimated NMSE of a channel estimate is below the (e.g., performance) threshold.

[0116] The WTRU may be configured to use non-specific performance monitoring allocations. The (e.g., non-specific performance monitoring) allocations may be immediate (e.g., signaled in DCI and / or similar control channel), and / or semi-static, for example with test allocation information indicated in SIB and / or other messages. The WTRU may receive the DMRS REs and / or apply some preprocessing, for example as shown in FIG. 6. Preprocessing may include one or more of extracting the DMRS REs from the received signals, division by and / or multiplication by conjugate of the known DMRS symbols, and / or a 2D filtering option of the DMRS REs. The WTRU may extract the DMRS REs from the received signals, for example usingthe DMRS resource allocation. The WTRU may be configured to apply the out-of- distribution (OOD) detection test and / or model drift detection test to the extracted DMRS REs, for example as described herein.

[0117] The WTRU may receive the data REs and / or apply preprocessing (e.g., FIG . 13), for example if the WTRU is configured to use type 2a and / or type 2b effective channel estimation performance monitoring allocations. Preprocessing may include one or more of extracting the data REs from the received signals using the data REs resource allocation, and / or division by and / or multiplication by conjugate of the known data REs symbols. Additionally, or alternatively, the WTRU may receive the L-DMRS and / or apply (e.g., some) preprocessing (e.g., as in FIG.6). Preprocessing may include one or more of extracting the L-DMRS REs from the received signals using the L-DMRS resource allocation, division by and / or multiplication by conjugate of the known L-DMRS symbols, and / or a 2D FIR filtering option of the L-DMRS carrying REs.

[0118] The WTRU may use the L-DMRS and / or the normal DMRS with a channel estimator that uses high-quality and quantity of input data to generate high-quality labels for the AI / ML-based channel estimator (e.g., as in FIG. 6). The WTRU may use the normal DMRS as input to the AI / ML-based channel estimator, for example if the WTRU is configured to use type 1a and / or type 1b effective channel estimation performance monitoring allocations. As shown FIG. 10, the predictions and / or outputs of AI / ML-based channel estimator may include one or more of the effective and / or precoded channel estimates at a (e.g., each) RE, the effective noise variance, and / or the Doppler estimate. The WTRU may use the normal DMRS and / or the data REs as input to the AI / ML-based channel estimator, for example if the WTRU is configured to use type 2a and / or type 2b effective channel estimation performance monitoring allocations. As shown in FIG. 6, the predictions and / or outputs of AI / ML-based channel estimator may include one or more of the effective and / or precoded channel estimates at each RE, the effective noise variance, and / or the Doppler estimate.

[0119] The WTRU may compute an error between the high-quality labels and the predictions of the AI / ML-based channel estimator. The WTRU may report the resulting error, for example to the gNB. The WTRU may compare the resulting error to a given threshold, for example configured by the gNB. The WTRU may be configured to use a (e.g., specific) error threshold for a (e.g., each) type of effective channel estimation performance monitoring allocation (e.g., type 1a, type 1b, type 2a, and / or type 2b). The error threshold may include a value between 0 and 1. The WTRU may continue to use the same AI / ML model, for example if the error is below the first threshold. The WTRU may request updating / retraining the AI / ML model while continuing to use the same AI / ML model, for example if the error is above the first threshold but below the second threshold. The WTRU may request retraining the AI / ML model while also using the fallback (e.g., non-AIML) algorithm for channel estimation for demodulation, for example if the error is above the second threshold.

[0120] Systems and methods for online learning of UL channel estimation for demodulation are disclosed herein. When a gNB monitors a performance degradation of an AI / ML model utilized for the uplink channel estimation for demodulation for example, the gNB may demand / request the support for retraining / updating the AI / ML model. The gNB may (e.g., then) retrain the AI / ML model online, for example with the assistance of another node in the network (e.g., WTRU). The online learning of an AI / ML-based uplink channel estimator for demodulation may enable modeladaptation to the local environments and to the radio impairments. The labels for AI / ML-based channel estimator may be created via one or more of new learning DMRS (L-DMRS) pilots, conventional NR RUSCH DMRS pilots (e.g., which may be referred to as DMRS pilots herein), and / or data symbols.

[0121] A WTRU may transmit the L-DMRS to generate the labels at a gNB for the trainable AI / ML-based uplink channel estimator for demodulation. The NW may detect a need for initiating online learning of uplink channel estimation for demodulation. The determination of the need for online learning may include certain WTRU (e.g., new type of WTRU and / or WTRU moving in a new way) and / or may be more general, for example related to a change in geography and / or weather.

[0122] Systems and methods are disclosed herein for AI / ML-based channel estimation for demodulation. The WTRU and / or the NW may perform the downlink and / or the uplink channel estimation for demodulation, for example via an AI / ML model. The AI / ML process may include one or more of an application, input data, preprocessing, AI / ML model, postprocessing, output data, and / or training. The application may include a channel estimation for demodulation. Input data may include received DMRS REs. The WTRU and / or the NW may perform preprocessing. Preprocessing may include one or more of extraction of the received DMRS REs from the received signals using the configured DMRS locations (e.g., allocations for the downlink and grants for the uplink), division of the received DMRS REs by (e.g., known) DMRS pilots, multiplication of the received DMRS REs by conjugate of the known DMRS pilots, applying a 2D filtering to the DMRS carrying REs, resizing the input data shape, and / or concatenation of the real and imaginary parts to obtain the real-valued input to the AI / ML model. AI / ML models may be for the channel estimation (e.g., residual channel estimation network (ReEsNet), ChannelNet, DDAE, and / or MTRE). The ReEsNet model may achieve higher channel estimation accuracy with low AI / ML model complexity (e.g., number of trainable parameters and / or computational complexity in floating point operations) compared to a benchmark (e.g., a non-AI / ML based channel estimation).

[0123] In the downlink for example, the WTRU may be configured to receive the L-DMRS for the online learning of AI / ML-based downlink channel estimator for demodulation based on the effective channel estimation training allocations (e.g., type 1 and / or type 2 in FIG. 7). Additionally, or alternatively, the WTRU may be configured to receive the L-DMRS for the performance monitoring of AI / ML-based downlink channel estimator for demodulation via the effective channel estimation performance monitoring allocations (e.g., type 1 and / or type 2 in FIG. 7). The effective channel estimation training and / or performance monitoring allocationsmay be immediate (e.g., signaled in RRC, DCI, MAC-CE, and / or similar control channels), and / or semi-static with training allocation information, for example indicated in SIB and / or other broadcast messages. The effective channel estimation training and / or performance monitoring allocations may be non-specific to any WTRU. Multiple WTRUs may use the same allocations, for example simultaneously. The non-specific semi-static allocations may be broadcast (e.g., by the network), for example so a (e.g., any) WTRU with capabilities to read SIB may read the effective channel estimation training and / or performance monitoring allocations.

[0124] In the uplink for example, the WTRU may be configured to transmit the L-DMRS based on the effective channel estimation grant (e.g., type 1 and / or type 2 in FIG. 7), The WTRU may be configured to transmit the L-DMRS based on the effective channel estimation grant when the NW detects a need for generating high-quality labels to be utilized in the online learning and / or the performance monitoring of the AI / ML-based uplink channel estimator for demodulation. The NW may configure the WTRU for the effective channel estimation grant via higher layer signaling, for example one or more of RRC, DCI, MAC-CE, and / or similar control channels.

[0125] FIG. 7 shows examples of type 1 and type 2 effective channel estimation training and / or performance monitoring allocations 700 in the downlink. The resource allocations described for type 1 and / or type 2 may be applicable for the effective channel estimation grant the in the uplink.

[0126] Herein, type 1 and / or type 2 may be described for the effective channel estimation training allocations in the downlink, but the descriptions may also applicable for the effective channel estimation performance monitoring allocations in the downlink and / or the effective channel estimation grant the in the uplink. Type 1 effective channel estimation training allocations may include L-DMRS, for example as in FIG. 7. In this allocation, none of the allocated RBs and / or RBGs carry the user data. The allocated RBs and / or RBGs may be populated with L-DMRS REs, for example for high-quality label generation during the AI / ML model training and / or performance monitoring. During the normal downlink data allocations for example there may be no L-DMRS, for example as in FIG. 8. The AI / ML model may use the normal density and / or low density DMRS REs as input(s). The allocated RBs and / or RBGs may also carry one or more of DMRS REs, CSI-RS REs, and / or control channels, for example PDCCH and / or other PHY channels. Type 1 effective channel estimation training allocations may be divided into (e.g., two) subcategories, for example as in FIG. 7. Type 1a effective channel estimation training allocations may include the allocated RBs and / or RBGs, which may carry normal density DMRS REs. Type 1 b effective channel estimation training allocations may include the allocated RBs and / or RBGs, which may carry low density DMRS REs. In type 1b forexample, a lower number of REs may be reserved for DM RS pilots compared to the DM RS patterns (e.g., 5G NR). The reduced DMRS overhead in type 1b may result in the allocation of more data REs in the RBs and / or RBGs (e.g., FIG. 8), which for example may increase the spectral efficiency. Remaining (e.g., most or all remaining) REs may be populated with L-DMRS REs, for example during the AI / ML model training and / or performance monitoring.

[0127] Type 2 effective channel estimation training allocations may include the L-DMRS and / or data, for example as in FIG. 7. The allocated RBs and / or RBGs may carry one or more of (e.g., both) the L-DMRS REs and / or data REs. The allocations and / or density of the L-DMRS REs and / or the data REs in the allocated RBs and / or RBGs may vary, for example based on predefined effective channel estimation training allocation configuration (e.g., in FIG. 7). In type 2a, 59 REs may be allocated for L-DMRS and / or 60 REs may be allocated for data in some examples. The density of L-DMRS REs and / or data REs may (e.g., then) be approximately 50%. During the normal downlink data allocations for example, there may be no L-DMRS (e.g., as in FIG. 8). The AI / ML model may use the normal density and / or low density DMRS REs and / or data REs as the input(s).

[0128] The allocated RBs and / or RBGs) may additionally, or alternatively carry DMRS REs, CSI-RS REs, and / or control channels (e.g., PDCCH), and / or other PHY channels. Type 2 may be divided into (e.g., two) subcategories, for example as in FIG. 7. Type 2a effective channel estimation training allocations may include the allocated RBs and / or RBGs, which for example may carry normal density DMRS REs. Type 2b effective channel estimation training allocations may include the allocated RBs and / or RBGs, which may carry low density DMRS REs. In type 2b, a lower number of REs may be reserved for DMRS pilots compared to the DMRS patterns (e.g., in 5G NR). The reduced DMRS overhead in type 2b may lead to the allocation of more data REs in the RBs and / or RBGs which, for example may increase the spectral efficiency. The remaining REs may be populated with the L-DMRS REs and / or the data REs, for example during the AI / ML model training and / or performance monitoring.

[0129] Systems and methods are disclosed for configurations for L-DMRS port structure(s). The L-DMRS structure may support (e.g., at least) the same number of antenna ports supported by DMRS. The L-DMRS may separate antenna ports using one or more of a similar time, frequency, and / or code domain orthogonalization techniques as in the DMRS. The L-DMRS may have higher density than the DMRS. The WTRU may be additionally, or alternatively be configured for the L-DMRS in the same RBs and / or RBGs, for example when the WTRU is configured for the normal density DMRS (e.g., NR PDSCH DMRS for downlink and / or NR PUSCH DMRS for uplink).

[0130] FIG. 9 shows an example 900 port structure 900 in one RB-slot for supporting 4 layers using antenna ports 1000-1003. The DMRS pattern may follow one or more of a single DMRS symbol, configuration type 1 , one additional position, mapping type A, and / or starting symbol 2. L-DMRS are (e.g., then) grouped into 10 groups, for example with the same pattern of the DMRS.

[0131] FIG. 10 shows an example port structure 1000 in one RB-slot for supporting 12 layers using antenna ports 1000-1011. The DMRS pattern may follow one or more of double DMRS symbol, configuration type 2, mapping type A, and / or starting symbol 2. L-DMRS may (e.g., then) be grouped into 5 groups, for example with the same pattern of the DMRS. The WTRU may be configured for the L-DMRS in the same RBs and / or RBGs, for example when the WTRU is additionally or alternatively configured for the low density DMRS. The L-DMRS may support higher antenna ports compared to the DMRS.

[0132] FIG. 11 shows an example port structure 1100 in one RB-slot, for example for the L- DMRS and / or low density DMRS to support a single layer. The DMRS pattern may follow one or more single DMRS symbol, mapping type A, and / or starting symbol 2. A lower number of REs are reserved for the DMRS compared to the DMRS configurations (e.g., configuration type 1 and / or type 2).

[0133] FIG. 12 shows an example port structure 1200 in one RB-slot, for example for the L- DMRS and / or low density DMRS to support 8 layers. The DMRS pattern may follow one or more of single DMRS symbol, mapping type A, and / or starting symbol 3. A lower number of REs may be reserved for the DMRS compared to the DMRS configurations (e.g., configuration type1 and / or type 2). Configuration type 1 in 5G NR may utilize double DMRS symbols (e.g., at least2 OFDM symbols) to support 8 layers. In FIG. 12 the DMRS may be configured for the ports 1000, 1002, 1004, 1006. There may be no DMRS REs for the ports 1001 , 1003, 1005, 1007. The L-DMRS configuration may cover all antenna ports 1000-1007 for generating the high- quality labels for each port. An AI / ML model trained with the high-quality labels may use the low density DMRS REs with the configured ports as the inputs and / or may (e.g., then) perform the predictions in the spatial-domain (e.g., predicting the ports 1001 , 1003, 1005, 1007), for example in addition to the time-domain and frequency-domain.

[0134] The variable L-DMRS RE and / or data RE patterns may be scheduled, for example so that over time a larger number of REs will carry data so that this larger number of REs can be used as input to the AI / ML-based channel estimator. The variable L-DMRS RE and / or data RE patterns may be scheduled for the type 2 effective channel estimation training allocations.

[0135] FIG. 13 shows an example resource allocation pattern 1300 for type 2 effective channel estimation training allocations, where for example L-DMRS is allocated in half of available REs and / or the data is allocated in the other half of available REs. Alternating slots may (e.g., then) be scheduled, where for example the set of REs used for the data in even slots may include the set of REs used for L-DMRS in odd slots. FIG. 13 shows an example of nested AI / ML model architecture as herein, which for example may include three sub-models (e.g., sub-model A, sub-model B, and / or sub-model C). The nested AI / ML model may use (e.g., all) data resources during inference. In odd slots for example, the sub-model A and / or the sub-model C may be trained. In even slots for example, the sub-model B and / or the sub-model C may be trained. During inference for example, all sub-models A, B, C may be active and / or all the data REs in the allocation may be used. Only the data RE inputs are shown for clarity. The sub-model C of the AI / ML model may not need to be trainable, for example it may average outputs of sub-model A and / or sub-model B.

[0136] When a WTRU indicates to the gNB that it supports type 2 allocations for example, the WTRU may additionally, or alternatively indicate that the WTRU supports variable pattern type 2 allocations. The number of versions of the pattern may be part of the semi-static allocation information. The initial version and a (e.g., each) subsequent version may be part of the semistatic allocation information. The pattern versions may be enumerated (e.g., 0, 1 , 2, ..., K-1). The initial pattern version slot number may be part of the semi-static allocation information. The WTRU may (e.g., then) determine (e.g., count) the number of slots used for training since the initial version slot number, for example to determine which pattern version is used on any other slot. The count may be modulo K. K may be either the number of versions and / or is provided as part of the semi-static allocation information. The pattern for each pattern version may be pre-determined and / or known to the WTRU (e.g., similar to how redundancy versions are known). Pattern versions for different rank transmission may be different, for example since the DM RS and / or L-DRMS patterns may additionally or alternatively be different as the rank or number of layers changes.

[0137] Systems and methods disclosed herein may include benefits, for example including the WTRU and / or the NW ability to generate the high-quality labels for online training and / or performance monitoring of AI / ML models utilized in the channel estimation for demodulation. The high-quality label generation may enhance the channel estimation quality, which for example may lead to improved overall system performance, and / or reduce the DM RS overhead size, which may lead to increased spectral efficiency.

[0138] Systems and methods for online learning of DL channel estimation for demodulation are disclosed herein. The WTRU and / or the NWAI / ML endpoint (e.g., a gNB, a cloud server, and / or an edge computing server) may detect and / or be informed of performance degradation of an AI / ML model that, for example requires action. When a WTRU (e.g., first) enters a (e.g., new) geographic region, the AI / ML model may not be well trained for this (e.g., new) local environment. The WTRU may request support for retraining / updating the AI / ML model. The WTRU may retrain the AI / ML model online, for example with the assistance of another node in the network (e.g., gNB, another WTRU using sidelink or WTRU-to-WTRU direct communication, and / or etc.). Online learning of an AI / ML-based downlink channel estimator for demodulation may (e.g., then) enable true adaptation to the local environments and / or the radio impairments. The labels for AI / ML-based downlink channel estimator may be created via new learning DMRS (L-DMRS) pilots and / or conventional NR PDSCH DMRS pilots. The WTRU may generate the labels for the downlink channel estimation and / or perform the online learning of the AI / ML-based downlink channel estimator.

[0139] Parameters (e.g., key parameters) related to the AI / ML functions may be transmitted from the WTRU to the NW, for example when a WTRU is configured to perform online learning of an AI / ML-based channel estimation for demodulation. The parameters may include one or more WTRU capabilities. The WTRU may be configured to use effective channel estimation training allocations and / or other training signals, for example depending on the WTRU capabilities. WTRU capabilities may include one or more of being configured to use type 1a effective channel estimation training allocations, being configured to use type 1 b effective channel estimation training allocations, being configured to use type 2a effective channel estimation training allocations for training without data REs, being configured to use type 2a effective channel estimation training allocations for training with data REs, being configured to use type 2b effective channel estimation training allocation without data REs, and / or being configured to use type 2b effective channel estimation training allocation with data REs. REs reserved for data may be used as input to the trainable channel estimator for demodulation, for example for type 2a effective channel estimation training allocations for training without data REs. The WTRU configured to support type 2a with data REs may include the data REs as input to the trainable channel estimator for demodulation. For example the WTRU may be configured to use type 2a effective channel estimation training allocations for training with data REs. If for example, the WTRU supports type 2a only without data REs, then WTRU may ignore the data REs as input to the trainable channel estimator for demodulation.

[0140] Other parameters (e.g., key parameters) related to the AI / ML functions may be configured by the NW, for example via higher layer signaling (e.g., RRC signaling, SIB, and / or etc.). The configuration may include one or more information elements and parameters including effective channel estimation training allocations scheduling, WTRU specific REs multiplexing with type 2 effective channel estimation training allocations, effective channel estimation training allocations resource mapping (e.g., antenna ports, effective channel estimation training allocations patterns, and / or etc.), high quality channel estimators (e.g., effective channel estimation training allocations modulation pattern), grouping of WTRUs for different effective channel estimation training allocations opportunities, size of training allocations, the set of usable RBs, and / or AI / ML model training parameters (e.g., loss function, training performance threshold, and / or etc.). Effective channel estimation training allocations may be non-specific to any (e.g., a specific) WTRU and / or non-specific to any WTRU within a WTRU L-DMRS group(s) indicated for the training allocation. Effective channel estimation training allocations may be configured with higher power (e.g., level) than normal DMRS and / or may be indicated in the training allocation as an offset from DMRS power (e.g., level).

[0141] WTRU specific REs may multiplex with type 2 effective channel estimation training allocations. For example, effective channel estimation training allocations may multiplex concurrently with data and DMRS REs transmitted in NR PDSCH (e.g., as well as other RS, such as, normal CSI-RS, and / or etc.). A parameter, RL-DMRS, which for example may be part of training allocation information, may be used to determine what ratio of the REs in the training allocation are used for L-DMRS. RL-DMRS 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) OL-DMRS and / or NL-DMRS may be used to determine the starting RE offset and / or number of effective channel estimation training allocations to be inserted in the type 2 allocation. Additionally, or alternatively, if the REs are listed (e.g., made ordinal), (e.g., then) the OL-DMRS may indicate the 1stRE to carry effective channel estimation training allocations. RL-DM S may be used to determine how many REs to skip until next effective channel estimation training allocations are inserted. NL-DMRS may be used to determine how many effective channel estimation training allocations are inserted and / or the RE number where effective channel estimation training allocations are no longer inserted. The ordinal list may omit a predetermined set REs such as DMRS. The predetermined list may be part of the training allocation information.

[0142] A WTRU may be configured to perform an OOD detection test, for example for an extracted DMRS. The DMRS may be input to the AI / ML-based channel estimation for demodulation. The (e.g., key) parameters may be configured by the NW via higher layersignaling, for example RRC signaling, SIB, and / or etc., as herein. Key parameters may be configured by the NW via higher layer signaling, for example RRC signaling, SIB, and / or etc., as herein. For example, the key parameters may be configured by the NW when a WTRU is configured to perform drift detection of the AI / ML-based channel estimation for demodulation. Additionally, or alternatively, the key parameters may be configured by the NW when a WTRU is configured to monitor the performance of the AI / ML-based channel estimation for demodulation via L-DMRS.

[0143] The configuration may include one or more effective channel estimation performance monitoring allocations in the performance monitoring of the channel estimation for demodulation. Additionally, or alternatively, the configuration may include test BW and / or a number of RBs to define how large the monitoring allocations will be. In some examples, the configuration may include a performance threshold (e.g., indication that the estimated NMSE of channel estimate is below a threshold to continue using the same AI / ML model). The effective channel estimation performance monitoring allocations in the performance monitoring of the channel estimation for demodulation may include one or more of the WTRU configured to use one or more of type 1 a, type 1 b, type 2a, and / or type 2b effective channel estimation performance monitoring allocations. Additionally, or alternatively, the effective channel estimation performance monitoring allocations may be immediate (e.g., signaled by DCI) and / or semi-static with test allocation information indicated in SIB and / or other broadcast message.

[0144] Systems and methods for signaling of online training are disclosed herein. A WTRU may utilize an online training request and / or online training response to initiate online training. The WTRU may transmit a message of an online training request, for example to the NW AI / ML end point node. The NW may decide a size of an effective channel estimation training allocation, for example based on the request.

[0145] The online training request may include one or more of a BW or number of RBs. Additionally, or alternatively, the online training request may include channel statistics measurements. The channel statistics measurements may include one or more of delay spread, Doppler, K-factor of LOS components, power delay profile (PDF), and / or etc. In some examples the online training request may include one or more of a channel diversification request (e.g., increased Delay Spread or Doppler), a number of layers, a maximum number of layers, a list of number of layers desired for training, a desired sequence of DMRS configurations, an indication of whether AI / ML-based channel estimator uses the data REs as inputs, and / or a priority of request. Additionally, or alternatively, the online training request may include an indication to select (e.g., an indication of how to select) precoders. The indication to select (e.g., how toselect) precoders may include one or more of round robin, a list of precoder table indices, and / or be based on a (e.g., latest) CSI report. The priority of request may include the WTRU indicating a high priority, for example if the performance of the AI / ML-based channel estimator is below a certain threshold.

[0146] The WTRU may receive the performance threshold from the gNB. For example, the network may indicate in SIB and / or RRC that an estimated NMSE of the effective channel estimate, which may be above a threshold, may be utilized for the WTRU to make a priority request. The NW may decide to not override data REs with effective channel estimation training allocations, and / or not to increase the effective channel estimation training allocations, for example if the priority is not high. The WTRU may be configured to indicate a low priority, for example if the request is for predictive maintenance. The quality of an AI / ML-based channel estimator may be tested by comparing the effective channel estimate, for example the AI / ML model predictions and / or the high-quality labels generated via the L-DMRS. Additionally, or alternatively, the online training request may include an indication of the AI / ML model performance, for example based on the comparison. In some examples, the online training request may include an indication that the AI / ML model performance has degraded compared to a threshold. The WTRU may signal the NW with a new message to increase the priority of training allocation, for example based on the comparison (e.g., the AI / ML model performance / degradation). The NW may be configured to receive a channel diversification request from the WTRU. The NW may be configured to determine a variation of the channel (e.g., a time-varying FIR filter may be used to augment the channel), for example based on the channel diversification request. For example, the NW may determine to add to the channel.

[0147] The NW may transmit the online training response message, for example at an AI / ML end point node. For example, the NW may transmit the online training response message in response to the online training request transmitted by the WTRU. The online training response may include one or more of the location of RBs or PRBs (e.g., where the L-DMRS is transmitted), the type of effective channel estimation training allocations (e.g., type 1a, type 1b, type 2a, type 2b, and / or etc.), an allocated online training duration (e.g., number of slots) and / or the starting time, an allocated online training L-DMRS pattern, port structure, density, and / or power offset, an indication of WTRU non-specific and / or specific effective channel estimation training allocation, and / or etc. The WTRU may transmit the online training request message on one or more of PUCCH, RRC, and / or NAS message.

[0148] A WTRU may be configured for the online training of an AI / ML model. The AI / ML model may be utilized in the channel estimation for demodulation. FIG. 14 shows an example of anAI / ML process and / or system 1400. In some examples, the AI / ML process and / or system may be used for channel estimation for demodulation. At 1402 the WTRU and / or network may determine if online training has started. At 1404 the WTRU may receive and / or learn a signaled effective channel estimation training allocations schedule. At 1406 the WTRU may receive DMRS REs and / or L-DMRS REs. For example, the AI / ML model input may include received DMRS REs. The DMRS REs may include normal density DMRS REs, for example for typela and / or type 2a (e.g., FIG. 7). Additionally, or alternatively, the DMRS REs may include low density DMRS REs, for example for typelb and / or type 2b (e.g., FIG. 7). For type 2 for example (e.g., FIG. 7), the received data REs may be additionally, or alternatively be used as the input data.

[0149] The AI / ML Model may be used for channel estimation (e.g., ReEsNet, ChannelNet, DDAE, and / or MTRE). The WTRU may postprocess by conversion of the real-valued AI / ML model output to the complex-valued output data. Additionally, or alternatively, the WTRU may postprocess by resizing the AI / ML model output. AI / ML model output may include one or more of effective and / or precoded channel estimates at a (e.g., each) RE, effective noise variance, and / or Doppler estimate.

[0150] The WTRU and / or the network may perform online training in a supervised manner. For example, the WTRU and / or the network may train by including features of one or more of DMRS REs and / or data REs (e.g., for type2). The WTRU and / or the network may generate labels online, for example via a high-quality-input-based channel estimator as in FIG. 6.

[0151] The WTRU may use the received L-DMRS REs and / or DMRS REs at the high-quality- input-based channel estimator. At 1408 the WTRU may apply preprocessing. The WTRU may apply preprocessing to the received L-DMRS REs and / or DMRS REs, for example before inputting the preprocessed received L-DMRS REs and / or DMRS REs into the high-quality-input- based channel estimator. The high-quality-input-based channel estimator may employ high- quality and / or quantity of input RSs. At 1410 the WTRU may use the high-quality-input-based channel estimator to generate the labels. The high-quality-input-based channel estimator may include one or more of the least square estimator, 2D filtering, up-sampling, denoising, interpolation, and / or etc.

[0152] The labels may include one or more of the effective and / or precoded channel estimates at each RE, the effective noise variance, and / or the Doppler estimate. The WTRU and / or the network may use the labels to train the AI / ML-based channel estimator. The received DMRS REs and / or the labels may be saved in a replay buffer, for example for a continued model training. The WTRU may save the labels in a replay buffer for performance monitoring.

[0153] At 1412 the WTRU and / or the network may input the high-quality labels and / or train the AI / ML based channel estimator. The WTRU and / or network may train the AI / ML-based channel estimator based on the generated labels and / or the predictions (e.g., outputs of AI / ML model), for example as in FIG. 6. The WTRU and / or network may monitor the training progress, for example by measuring the rate of change of the loss function and / or comparing to a threshold. The WTRU may determine whether the online training is complete at 1414. For example, the WTRU may determine that training is complete / sufficient when the loss function has a rate of change less than a threshold. If the WTRU determines that online training is not complete the procedure 1400 may return to 1406. For example, the WTRU may continue training and / or not send a message to the network if the WTRU determines that training is not complete / sufficient. At 1416, the WTRU may send a training complete message to the gNB, for example when the WTRU determines (e.g., detects) that training is sufficient. The gNB may stop transmitting and / or allocating L-DMRS REs (e.g., as in FIG. 8), for example upon receiving the training complete message.

[0154] At 1418 the WTRU may receive DMRS REs and / or data REs (e.g., type 2). The WTRU may apply preprocessing at 1420, for example on the DMRS REs and / or data REs (e.g., type 2). At 1422 preprocessed DMRS REs and / or data REs may be input into the AI / ML based channel estimator and / or the WTRU may use the AI / ML based channel estimator to generate effective channel estimates. The WTRU and / or AI / ML based channel estimator may output the effective channel estimates. At 1422 the WTRU may determine whether online training has started, for example based on the output effective channel estimates. If the WTRU determines that online training has not started, the WTRU may output the effective channel estimates on its own and / or use the effective channel estimates in demodulation. If the WTRU determines that online training has started, the WTRU may output the effective channel estimates as input for training the AI / ML based channel estimator, for example at 1412. In some examples, the AI / ML based channel estimator may output an AI / ML based channel estimator update, for example generating effective channel estimates at 1422.

[0155] The WTRU may make an inference. For example, the WTRU may use the trained AI / ML-based channel estimator for demodulation. The WTRU may monitor performance. The WTRU may perform the OOD detection test to the extracted DMRS as described herein. Additionally, or alternatively, the WTRU may perform the AI / ML model drift detection as described herein. In some examples, the WTRU may perform the performance monitoring of the AI / ML-based channel estimation via L-DMRS as described herein, for example during the inference.

[0156] The AI / ML model may be used for training and / or high-quality label generation. The WTRU and / or the NW may detect a need for initiating the online learning of channel estimation for demodulation. The WTRU may utilize a system or method for AI / ML performance monitoring via L-DMRS as described herein. The NW may detect that the performance of channel estimation for demodulation may be degraded, for example based on existing ACK / NACK signaling and / or CSI reporting. The gNB may signal the WTRU with a message (e.g., via MAC CE and / or similar signaling) that performance has degraded, for example if the gNB uses the indicated PMI and / or MCS for a WTRU over several transmissions and / or the gNB receives a larger than expected number of NACKs within a time window.

[0157] Expected NACKs may include adjustment to the CQI, for example to account for historical bias from each WTRU as estimated by the gNB. The message (e.g., from the gNB to the WTRU) may include a notification of suspected degradation and / or instruction to test channel estimation for demodulation performance. Additionally, or alternatively, the message may include an instruction to request online training of channel estimation for demodulation and / or an instruction to use the effective channel estimation training allocations already configured by the network. The notification of suspected degradation and / or instruction to test channel estimation for demodulation performance may include the DM RS configuration(s) with suspected degradation (e.g., number of layers, and / or DMRS density).

[0158] The WTRU may request online training of channel estimation for demodulation as described herein. The WTRU may use the effective channel estimation training allocations as described herein. The WTRU may receive the information (e.g., from the network), for example to learn the current training allocation schedules. The WTRU may read the information (e.g., SIB) or RRC signaling, for example for learning the semi-static schedule of effective channel estimation training allocations. The information may include one or more of the L-DMRS REs allocation information and / or the L-DMRS power offset information.

[0159] The WTRU may be assigned to an L-DMRS group by the network. An L-DMRS group / grouping may be utilized in directional systems, for example where different ports may be used for different directions (e.g., groups of WTRUs). WTRUs with similar CSI reports may be grouped under the same L-DMRS group to share the effective channel estimation training allocations with similar precoding, for example based on the CSI reports. Semi-static training allocations, for example within a group, may be non-specific (e.g., any WTRU in the group(s) may simultaneously use the training allocation). The associated WTRU group or groups may be indicated in the training allocation information in the SIB, for example if multiple effective channel estimation training allocations are used for multiple groups. Training allocations mayadditionally, or alternatively be made dynamically (e.g., via control channel, PDCCH / DCI), where for example the WTRU group or groups may be indicated in the dynamic control information. The WTRU may read the schedule for the L-DMRS group that it belongs to and / or (e.g., only) use the corresponding schedules for AI / ML model training and performance monitoring.

[0160] FIG. 6 shows a WTRU configured to receive the DM RS REs and / or apply the preprocessing. The WTRU may receive the data REs and / or apply the preprocessing, for example for type 2. The WTRU may use the received L-DMRS REs and / or the DMRS REs at the high-quality-input-based channel estimator, for example during the AI / ML model training and / or performance monitoring. In some examples, the WTRU may use the received DMRS REs at the AI / ML-based channel estimator. The WTRU may additionally, or alternatively use the received data REs at the AI / ML-based channel estimator, for example for the type 2 training allocation (e.g., FIG. 7).

[0161] The WTRU may perform different preprocessing on the received DMRS REs for the high-quality-input-based channel estimator and AI / ML-based channel estimator. The WTRU may train the AI / ML-based channel estimator based on the generated high-quality labels and / or the prediction, for example as in FIG. 6. The WTRU may monitor the training progress. When the WTRU detects that training is sufficient for example, the WTRU may send a training complete message to the gNB. The gNB may stop transmitting and / or allocating L-DMRS REs (e.g., as in FIG. 8), for example when the training complete message is received. The WTRU may use the trained AI / ML-based channel estimator for online inference.

[0162] Systems and methods for WTRU AI / ML performance monitoring via L-DMRS are disclosed herein. FIG. 15 shows an example procedure 1500 for the WTRU to monitor the performance of the AI / ML-based channel estimation. The WTRU may be configured to apply one or more of the following, for example as in FIG. 6 (e.g., for the type 2 training allocation). At 1502 the WTRU may apply preprocessing to the received L-DMRS REs and / or DMRS REs, for example before forwarding the preprocessed received L-DMRS REs and / or DMRS REs to the high-quality-input-based channel estimator. At 1504 the WTRU may apply the preprocessed L- DMRS REs and / or DMRS REs at the high-quality-input-based channel estimator, for example to generate high quality labels. The high-quality-input-based channel estimator may use the high- quality and / or quantity input RSs to generate high-quality labels for the AI / ML-based channel estimator. The WTRU may use the normal DMRS as input to the AI / ML-based channel estimator, for example if the WTRU is configured to use type 1a and / or type 1b effective channel estimation performance monitoring allocations. Additionally, or alternatively, the WTRUmay use the normal DMRS and / or the data REs as input to the AI / ML-based channel estimator, for example if the WTRU is configured to use type 2a and / or type 2b effective channel estimation performance monitoring allocations. At 1506 the WTRU may apply the preprocessed DMRS REs and / or data REs at the AI / ML based channel estimator, for example to generate effective channel estimates.

[0163] At 1508 the WTRU may compute the error between the high-quality labels and the effective channel estimates (e.g., predictions of the AI / ML-based channel estimator). The predictions and / or outputs of AI / ML-based channel estimator may include one or more of the effective and / or precoded channel estimates at a (e.g., each) RE, the effective noise variance, and the Doppler estimate. At 1510 the WTRU may send an error report (e.g., report the resulting error), for example to the gNB. In some examples, 1510 may be omitted. At 1512 the WTRU determine if the error is below a (e.g., configured) second threshold, for example by comparing the resulting error to a given threshold configured by the gNB. The WTRU may be configured to use a specific error threshold for each type of effective channel estimation performance monitoring allocation (e.g., type 1a, type 1 b, type 2a, and / or type 2b).

[0164] The procedure 1500 may proceed to 1514 if the WTRU determines that the error is below the (e.g., configured) second threshold at 1512. At 1514 the WTRU may determine if the error is below a (e.g., configured) first threshold, for example by comparing the resulting error to a given threshold configured by the gNB. If the error is below a first threshold the WTRU may continue to use the same AI / ML model (e.g., channel estimator) at 1516. If the error is above the first threshold but below the second threshold, the WTRU may request updating / retraining the AI / ML model at 1518 (e.g., from the network), for example while continuing to use the same AI / ML model (e.g., channel estimator at 1518). In some examples, the WTRU requesting updating / retraining the AI / ML model at 1518 may be omitted. If at 1512 the error is above the second threshold the WTRU may request updating / retraining the AI / ML model at 1520. In some examples, the WTRU requesting updating / retraining the AI / ML model at 1520 may be omitted. At 1522 the WTRU may use a fallback (e.g., non-AIML) channel estimator, for example for demodulation. At 1524 the WTRU may estimate the effective channel at (e.g., each) RE associated with the (e.g., received) DMRS.

[0165] Example numerical results are disclosed herein. The numerical results may be presented for the performance evaluation of the proposed online learning of the channel estimation for demodulation. The simulations may be performed via an open-source Python library for the link-level simulations based on TensorFlow, for example Sionna. Example simulation parameters are shown in Table 1.Table 1. Example simulation parameters

[0166] FIG. 16 is an example OFDM resource grid 1600 with L-DMRS and DMRS allocations. In the OFDM resource grid, the low-density DMRS REs may be located at the 2ndand 11thOFDM Symbols with the frequency spacing (e.g., interval) of 6 REs as illustrated in FIG. 16. Selected low-density DMRS patterns for some simulations, for example as in FIG. 16, may not be supported by 5G NR. Available REs (e.g., all available REs) in the allocated 11 RBs may be populated with L-DMRS REs (e.g., no data REs) and / or may include an example of type 1b effective channel estimation training allocation (e.g., FIG 7), for example during online training. The OFDM resource grid with 14 OFDM symbols and 132 subcarriers may include 1848 REs with the allocation of 44 DMRS REs and / or 1804 L-DMRS REs. The L-DMRS REs may have higher density compared to the DMRS REs. Additionally, or alternatively, the power of DMRS REs and L-DMRS REs may be the same during the simulations. Channel normalization may be enabled in Sionna during the simulations.

[0167] For an example dataset, 20.000 samples may be generated for each SNR level. In the example of Table 1 , the dataset may include a total of 100.000 samples. The samples may be split into 60% for training, %20 for validation, and / or 20% for testing. The AI / ML-based channel estimator may be trained with Adam optimizer, for example with a learning rate of 0.001 , a batch size of 300, and / or a total epoch of 100. The target of an AI / ML-based channel estimator maybe to predict the effective channel estimates at each RE, for example via the received DM RS REs as the input data.

[0168] An AI / ML process and / or system as herein may include one or more of preprocessing, an AI / ML model, postprocessing, output data, and / or training. Preprocessing may include one or more of the multiplication of the received DMRS REs by conjugate of the known DMRS pilots, resizing the input data shape, and / or concatenation of the real and imaginary parts, for example to obtain the real-valued input to the AI / ML model. In some examples, the AI / ML model input size may be 2 x 22 x 2. The first dimension (e.g., 2) may refer to the number of OFDM symbols with DMRS REs, the second dimension (e.g., 22) may refer to the number of subcarriers with DMRS REs, and / or the third dimension (e.g., 2) may refer to the concatenation of the real and imaginary parts.

[0169] The WTRU and / or the network may construct an AI / ML model based on the ReEsNet architecture, for example as in FIG. 5. For example, the AI / ML model output size may be 14 x 132 x 2. The AI / ML model may include approximately 52k parameters. Additionally, or alternatively, the AI / ML model may include one or more of convolutional and / or up-sampling layers. The WTRU and / or the network may postprocess, for example to convert the real-valued AI / ML model output to complex-valued output data. Output data may correspond to the effective channel estimates, for example at each RE with a size of 14 x 132. The WTRU and / or the network may perform training online using high-quality labels generated via L-DMRS REs and / or DMRS REs. The high-quality-input-based channel estimator may be implemented via least square estimation and / or filtering, for example for label generation.

[0170] FIG. 17 is an example channel estimation normalized mean square error (NMSE) 1700 for AI / ML-based and non-AI / ML-based channel estimators. The (e.g., proposed) AI / ML-based channel estimator may be compared with a non-AI / ML baseline (e.g., the least square estimator and linear interpolation). The numerical results may show that the AI / ML-based channel estimator significantly improves the channel estimation quality and / or achieves a lower NMSE compared to the non-AI / ML baseline. For example, when SNR is 0 dB, an AI / ML-based channel estimator may improve the NMSE of channel estimation by 7.4 dB. Additionally, or alternatively the gap in NMSE curves may increase to 9 dB, for example when SNR is 20 dB. The (e.g., proposed) AI / ML-based channel estimator may achieve a target NMSE even in lower SNR regimes. For example, if the target NMSE is defined as -16 dB, the (e.g., proposed) AI / ML scheme SNR may be at least 5 dB. Additionally, or alternatively, the minimum SNR requirement may increase to 15 dB for the non-AI / ML baseline

[0171] Systems and methods for online learning of UL channel estimation for demodulation are disclosed herein. A gNB may monitor a performance degradation of an AI / ML model utilized for the uplink channel estimation for demodulation. The gNB may demand (e.g., send a request for) retraining and / or updating the AI / ML model (e.g., to the WTRU). The gNB may (e.g., then) retrain the AI / ML model online, for example with the assistance of another node in the wireless communication system (e.g., WTRU). The online learning of an AI / ML-based uplink channel estimator for demodulation may enable model adaptation, for example to the local environments and / or to the radio impairments. The gNB / the network may create the labels for the AI / ML- based channel estimator via one or more of new learning DM RS (L-DMRS) pilots, conventional NR PUSCH DMRS pilots, and / or data symbols. The gNB and / or a WTRU may transmit the labels for the trainable AI / ML-based uplink channel estimator for demodulation.

[0172] Configurations are disclosed herein. A gNB may configure the WTRU to transmit L- DMRS and / or other training signals, for example for the gNB to perform online learning and / or performance monitoring of an AI / ML-based channel estimation for demodulation. The WTRU may transmit (e.g., key) parameters related to the AI / ML functions, for example to the network. The (e.g., key) parameters may include one or more WTRU capabilities. The WTRU may support the transmission of L-DMRS and / or other training signals, for example depending on WTRU capabilities and associated grant resource allocations. WTRU capabilities may include one or more of support for type 1a L-DMRS effective channel estimation uplink grant, support for type 1b L-DMRS effective channel estimation uplink grant, support for type 2a L-DMRS effective channel estimation uplink grant without data REs, support for type 2a L-DMRS effective channel estimation uplink grant with data REs, support for type 2b L-DMRS effective channel estimation uplink grant without data REs, and / or support for type 2b L-DMRS effective channel estimation uplink grant with data REs.

[0173] The network may configure other (e.g., key) parameters related to the L-DMRS and / or other training signals, for example via higher layer signaling. For example, the higher layer signaling may include one or more of RRC signaling of resource allocation for uplink transmission with configured grant, RRC (typel), PDCCH (e.g., addressed to CS-RNTI) (e.g., type2), SIB, and / or etc. The configuration may include one or more information elements and / or parameters. For example, the WTRU may receive a configuration that one or more of DCI signaling on PDCCH will be used to activate / deactivate the L-DMRS configured grant. The WTRU may receive L-DMRS resource mapping (e.g., antenna ports, L-DMRS patterns, and / or etc.), high quality channel estimators (e.g., L-DMRS modulation pattern), a grouping of WTRUs for different L-DMRS opportunities, a size of training allocations, a set of usable RBs, L-DMRSeffective channel estimation training allocations uplink configured grants, and / or WTRU data REs multiplexing with L-DMRS in type 2 training allocations. L-DMRS may be configured with a higher power level than normal DM RS and / or may be indicated in the training allocation, for example as an offset relative to the DM RS power level.

[0174] WTRU data REs may multiplex with L-DMRS in type 2 training allocations. For example, L-DMRS may be multiplexed concurrently with one or more of data, DMRS REs transmitted in NR PUSCH, and / or other RS (e.g., SRS, etc.) A parameter, RL-DMRS, which may be part of the training allocation information, may be used to determine a ratio of the REs in the training allocation to use for L-DMRS. RL-DMRS 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) OL-DMRS and / or NL-DMRS may be used to determine the starting RE offset and / or number of L-DMRS to be inserted in the type 2 allocation. For example, if the REs are listed (e.g., made ordinal), (e.g., then) OL-DMRS may indicate the 1stRE to carry L-DMRS. RL-DMRS may be used to determine how many REs to skip until next L-DMRS is inserted. NL-DMRS may be used to determine how many L-DMRS are inserted and / or an RE number where L-DMRS are no longer inserted. The ordinal list may omit a predetermined set REs, such as DMRS for example. The predetermined list may be part of the training allocation information.

[0175] Systems and methods for label generation for online learning and / or performance monitoring are disclosed herein. The network may detect a need for initiating online learning of uplink channel estimation for demodulation. The network may determine the need for online learning based on a certain WTRU (e.g., new type of WTRU and / or WTRU moving in a new way). Additionally, or alternatively, the network may determine the need for online learning based on a change in geography and / or weather. For cell wide training for example, a special semi-persistent and / or opportunistic grant may be signaled as broadcast (e.g., on SIB). WTRUs configured to support online uplink channel estimation for demodulation may be granted RB and / or slots and / or portions of slots orthogonal to each other (e.g., by the network). The grant information may indicate the pattern of precoders to use and / or a rank that should be used for L- DMRS. The precoder pattern may include a cycle thru a predefined precoder table. Additionally, or alternatively, the network may configure the WTRU to estimate the best precoder. The network may configure the WTRU with aperiodic grants for L-DMRS and / or with periodic semi-persistent grants in a similar way as for cell wide training, for example if the gNB determines to train the WTRU. A data grant to a (e.g., any) WTRU may preempt a grant, for example an opportunistic grant. A WTRU, with a semi-persistent training grant for example, may listen to DCI preceding the UL L-DMRS grant to determine of the opportunistic grant is stillavailable. The DCI may include an indication of whether the opportunistic grant is still available, for example in a field.

[0176] A WTRU, for example scheduled to participate in semi-persistent training grants, may signal the gNB to request exclusion from participation. A WTRU (e.g., scheduled to participate in semi-persistent training grants) may participate in a predetermined minimum number, K, of slots of training grants and / or (e.g., then) stop participating, for example until the SIB info for the training grants is changed. The minimum number, K, may be signaled in the SIB.

[0177] The gNB may configure the WTRU for online training of uplink effective channel estimation for demodulation. The WTRU may be configured to transmit L-DMRS. The WTRU may be configured with the effective channel estimation training grants. Type 1 (e.g., type 1a and / or type 1b) effective channel estimation training grants may include L-DMRS, for example as in FIG. 7. Type 2 (e.g., type 2a and / or type 2b) effective channel estimation training grants may include L-DMRS and / or data, for example as in FIG. 7. In type 1a and / or type 2a for example, the allocated RBs and / or RBGs may carry normal density DM RS REs. In type 1b and / or type 2b for example, the allocated RBs and / or RBGs may carry low density DM RS REs.

[0178] The WTRU may receive the information (e.g., from the network) to learn the current schedules for the effective channel estimation training grant. The WTRU may read the information (e.g., SIB and / or RRC) for learning the semi-static schedule of effective channel estimation training grants of one or more of type 1a, type 1b, type 2a, and / or type 2b. The WTRU may receive DCI signaling on PDCCH to activate / deactivate the L-DMRS configured grant. As shown in FIG.18 for example, the WTRU may transmit one or more of normal density DMRS REs (e.g., type 1a and / or type 2a), low density DMRS REs (e.g., type 1 b and / or type 2b), L-DMRS REs, and / or data REs (e.g., for type 2a and / or type 2b).

[0179] FIG. 18 is an example of a procedure 1800 for label generation and training and performance monitoring of uplink channel estimation. The gNB may (e.g., then) apply (e.g., some) preprocessing for high-quality label generation for online training and / or for performance monitoring. The WTRU may receive a training complete and / or a performance monitoring complete message from the gNB. Additionally, or alternatively, the WTRU may receive (e.g., from the gNB) an indication to stop and / or deactivate the transmission of L-DMRS and / or other training signals, for example via DCI and / or MAC-CE.

[0180] At 1802 the WTRU may send one or more L-DMRS REs and / or DMRS REs to a gNBThe DMRS REs may be associated with NR PUSCH DMRS. Additionally, or alternatively, the WTRU may send data to the gNB. At 1804 the gNB may train an AI / ML model for channel estimation. For example, the gNB may preprocess the one or more L-DMRS REs, DMRS REs,and / or data. In some examples, preprocessing may be omitted. The L-DMRS REs may be input into a high-quality-input-based channel estimator, for example after preprocessing. Additionally, or alternatively, one or more of an L-DMRS pattern and / or DM RS pattern may be preprocessed and / or input into the high-quality-input-based channel estimator. The high-quality-input-based channel estimator may generate labels, for example from one or more of the L-DMRS REs, DMRS REs, the L-DMRS pattern and / or the DMRS pattern. The labels may include one or more of effective channel estimates (e.g., at each RE) and / or noise variance. The gNB / network may monitor the training progress, for example by measuring the rate of change of the loss function and / or comparing to a threshold. A loss function may be used as input to the high-quality-input- based channel estimator, for example during training.

[0181] Additionally, or alternatively, at 1804 the gNB may input DMRS REs and / or data into an AI / ML-based channel estimator, for example after preprocessing. Additionally, or alternatively, one or more of an L-DMRS pattern and / or DMRS pattern may be preprocessed and / or input into the AI / ML-based channel estimator. The AI / ML-based channel estimator may generate predictions, for example from one or more of the DMRS REs, the data, the L-DMRS pattern and / or the DMRS pattern. The predictions may include one or more of effective channel estimates (e.g., at each RE) and / or noise variance. The gNB / network may monitor the training progress, for example by measuring the rate of change of the loss function and / or comparing to a threshold. A loss function may be used as input to the AI / ML-based channel estimator, for example during training.

Claims

CLAIMS:

1. A wireless transmit / receive unit (WTRU) comprising a processor, the processor configured to: receive channel estimation training allocation information; receive learning demodulation reference signal (L-DMRS) resource elements (REs) and DMRS REs according to the channel estimation training allocation information, wherein the L- DMRS REs comprise a first density and a first power level, and the DMRS REs comprise a second density and a second power level; generate labels based on the L-DMRS REs and the DMRS REs; train an artificial intelligence / machine learning (AIZML)-based channel estimator based on the generated labels; determine whether online training is complete based on the training of the AI / ML-based channel estimator; and send, based on the determination of whether online training is complete, an indication of whether online training is complete to a network.

2. The WTRU of claim 1, wherein first density is greater than the second density, and the first power level is greater than the second power level.

3. The WTRU of claim 1, wherein the processor is configured to determine whether online training is complete based on an error associated with the L-DMRS REs and the DMRS REs.

4. The WTRU of claim 3, wherein the processor is configured to determine that online training is not complete if the error is greater than a threshold, and wherein the indication of whether online training is complete comprises a request for retraining.

5. The WTRU of claim 3, wherein the processor is configured to: determine that online training is complete if the error is less than a first threshold and less than a second threshold, and wherein the indication of whether online training is complete comprises an indication that online training is complete.

6. The WTRU of claim 1, wherein the processor is configured to determine to initiate online learning of channel estimation for demodulation.

7. The WTRU of claim 6, wherein the processor is configured to determine whether to initiate online learning of channel estimation for demodulation based on a location change.

8. The WTRU of claim 1, wherein the processor is configured to send a request for the channel estimation training allocation information.

9. The WTRU of claim 1, wherein the processor is configured to determine whether online training is complete based on a rate of change of a loss function being less than a threshold.

10. The WTRU of claim 1, wherein the channel estimation training allocation information comprises one or more of L-DMRS RE allocation information or L-DMRS power offset information.

11. A method performed by a wireless transmit / receive unit (WTRU, the method comprising: receiving channel estimation training allocation information; receiving learning demodulation reference signal (L-DMRS) resource elements (REs) and DMRS REs according to the channel estimation training allocation information, wherein the L-DMRS REs comprise a first density and a first power level, and the DMRS REs comprise a second density and a second power level; generating labels based on the L-DMRS REs and the DMRS REs; training an artificial intelligence / machine learning (AI / ML)-based channel estimator based on the generated labels; determining whether online training is complete based on the training of the AI / ML-based channel estimator; and sending, based on the determination of whether online training is complete, an indication of whether online training is complete to a network.

12. The method of claim 11, wherein first density is greater than the second density, and the first power level is greater than the second power level.

13. The method of claim 11 , further comprising determining whether online training is complete based on an error associated with the L-DMRS REs and the DMRS REs.

14. The method of claim 13, comprising determining that online training is not complete if the error is greater than a threshold, and wherein the indication of whether online training is complete comprises a request for retraining.

15. The method of claim 13, comprising determining that online training is complete if the error is less than a first threshold and less than a second threshold, and wherein the indication of whether online training is complete comprises an indication that online training is complete.

16. The method of claim 11 , further comprising determining to initiate online learning of channel estimation for demodulation.

17. The method of claim 16, wherein determining whether to initiate online learning of channel estimation for demodulation is based on a location change.

18. The method of claim 11 , further comprising sending a request for the channel estimation training allocation information.

19. The method of claim 11 , further comprising determining whether online training is complete based on a rate of change of a loss function being less than a threshold.

20. The method of claim 11, wherein the channel estimation training allocation information comprises one or more of L-DMRS RE allocation information or L-DMRS power offset information.

Images