Methods, architectures, apparatuses and systems for federated learning of a nested artificial intelligence (AI) / machine learning (ML) model in wireless protocols

Abstract

Procedures, methods, architectures, apparatuses, systems, devices, and computer program products directed to a wireless transmit / receive unit (WTRU) which may be configured for sending, to a network node, first information indicating one or more capabilities and / or one or more conditions associated with the first WTRU; receiving, from the network node, based on the first information, configuration information for training a nested artificial intelligence / machine learning (AI / ML) model, wherein the nested AI / ML model comprises a plurality of AI / ML sub-models; performing, based on the configuration information, a training of at least one AI / ML sub-model from the plurality of AI / ML sub-models; and sending, to the network node, second information indicating any of: (1) a first set of model parameters associated with the training of the at least one AI / ML sub-model, and (2) a number of training iterations associated with the training of the at least one AI / ML sub-model.

Description

METHODS, ARCHITECTURES, APPARATUSES AND SYSTEMS FOR FEDERATED LEARNING OF A NESTED ARTIFICIAL INTELLIGENCE (Al) / MACHINE LEARNING (ML) MODEL IN WIRELESS PROTOCOLSCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 539,862 filed September 22, 2023, which is incorporated herein by reference in its entirety.TECHNICAL FIELD

[0002] The present disclosure is generally directed to the fields of communications, software and encoding, including, for example, to methods, architectures, apparatuses, systems directed to artificial intelligence and / or machine learning, for example to methods, apparatus, and systems for federated learning of nested AI / ML models.BRIEF DESCRIPTION OF THE DRAWINGS

[0003] A more detailed understanding may be had from the detailed description below, given by way of example in conjunction with drawings appended hereto. Figures in such drawings, like the detailed description, are examples. As such, the Figures (FIGs.) and the detailed description are not to be considered limiting, and other equally effective examples are possible and likely. Furthermore, like reference numerals ("ref.") in the FIGs. indicate like elements, and wherein:

[0004] FIG. 1 A is a system diagram illustrating an example communications system;

[0005] FIG. IB is a system diagram illustrating an example wireless transmit / receive unit (WTRU) that may be used within the communications system illustrated in FIG. 1 A;

[0006] 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;

[0007] FIG. ID is a system diagram illustrating a further example RAN and a further example CN that may be used within the communications system illustrated in FIG. 1 A;

[0008] FIG. 2 illustrates an example of federated learning over a 5G system;

[0009] FIG. 3 depicts an example of federated learning protocol over wireless communication systems;

[0010] FIG. 4 illustrates an example of a nested AI / ML model architecture;

[0011] FIG. 5 illustrates an example of variations of nested AI / ML model architectures based on input data size and layers;

[0012] FIG. 6 illustrates an example of variation of nested AI / ML model architectures based on the bit-depth;

[0013] FIG. 7 illustrates an example of variation of two-sided nested AI / ML model architectures;

[0014] FIG. 8 illustrates an example of variation of nested AI / ML model architectures with local models;

[0015] FIG. 9 is a diagram illustrating a signal flow for the distributed / federated training of nested AI / ML model architecture;

[0016] FIG. 10 illustrates an example of federated learning for a nested AI / ML model architecture;

[0017] FIG. 11 is a diagram illustrating a signal flow for the hierarchical federated training of nested AI / ML model architecture;

[0018] FIG. 12 illustrates an example of hierarchical federated learning for a nested AI / ML model architecture;

[0019] FIG. 13 depicts an example procedure for federated learning of nested AI / ML models implemented by a wireless transmit / receive unit (WTRU);

[0020] FIG. 14 depicts another example procedure for federated learning of nested AI / ML models implemented by a WTRU; and

[0021] FIG. 15 depicts another example procedure for federated learning of nested AI / ML models implemented by a network node.DETAILED DESCRIPTION

[0022] In the following detailed description, numerous specific details are set forth to provide a thorough understanding of embodiments and / or examples disclosed herein. However, it will be understood that such embodiments and examples may be practiced without some or all of the specific details set forth herein. In other instances, well-known methods, procedures, components and circuits have not been described in detail, so as not to obscure the following description. Further, embodiments and examples not specifically described herein may be practiced in lieu of, or in combination with, the embodiments and other examples described, disclosed or otherwise provided explicitly, implicitly and / or inherently (collectively "provided") herein. Although various embodiments are described and / or claimed herein in which an apparatus, system, device, etc., and / or any element thereof carries out an operation, process, algorithm, function, etc., and / or any portion thereof, it is to be understood that any embodiments described and / or claimed herein assume that any apparatus, system, device, etc., and / or any element thereof is configured to carry out any operation, process, algorithm, function, etc. and / or any portion thereof.

[0023] Example Communications System

[0024] The methods, apparatuses and systems provided herein are well-suited for communications involving both wired and wireless networks. An overview of various types of wireless devices and infrastructure is provided with respect to FIGs. 1A-1D, where variouselements of the network may utilize, perform, be arranged in accordance with and / or be adapted and / or configured for the methods, apparatuses and systems provided herein.

[0025] FIG. 1A is a system 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), singlecarrier FDMA (SC-FDMA), zero-tail (ZT) unique-word (UW) discreet Fourier transform (DFT) spread OFDM (ZT UW DTS-s OFDM), unique word OFDM (UW-OFDM), resource block- filtered OFDM, filter bank multicarrier (FBMC), and the like.

[0026] As shown in FIG. 1A, the communications system 100 may include wireless transmit / receive units (WTRUs) 102a, 102b, 102c, 102d, a radio access network (RAN) 104 / 113, a core network (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 wireless signals and may include (or be) 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 (e.g., UE).

[0027] 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, e.g., to facilitate access to one or more communication networks, such as the CN 106 / 115, the Internet 110, and / orthe networks 112. By way of example, the base stations 114a, 114b may be any of a base transceiver station (BTS), a Node-B (NB), an eNode-B (eNB), a Home Node-B (HNB), a Home eNode-B (HeNB), a gNode-B (gNB), a NR Node-B (NR NB), 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.

[0028] 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 an 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 or any sector of the cell. For example, beamforming may be used to transmit and / or receive signals in desired spatial directions.

[0029] 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).

[0030] 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 116 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 Packet Access (HSDPA) and / or High-Speed Uplink Packet Access (HSUPA).

[0031] 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).

[0032] 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).

[0033] 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., an eNB and a gNB).

[0034] In an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement radio technologies such as IEEE 802.11 (i.e., Wireless Fidelity (Wi-Fi), IEEE 802.16 (i.e., Worldwide Interoperability for Microwave Access (WiMAX)), CDMA2000, CDMA2000 IX, 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.

[0035] 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 an 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 an embodiment, the base station 114b and the WTRUs 102c, 102d may utilize a cellular-based RAT (e.g., WCDMA, CDMA2000, GSM, LTE, LTE-A, LTE-A Pro, NR, etc.) to establish any of a small cell, picocell or femtocell. As shown in FIG. 1 A, 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.

[0036] 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 varyingquality 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. 1 A, 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 an NR radio technology, the CN 106 / 115 may also be in communication with another RAN (not shown) employing any of a GSM, UMTS, CDMA 2000, WiMAX, E-UTRA, or Wi-Fi radio technology.

[0037] 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 other networks 112. The PSTN 108 may include circuit-switched telephone networks that provide plain old telephone service (POTS). The Internet 110 may include a global system of interconnected computer networks and devices that use common communication protocols, such as the transmission control protocol (TCP), user datagram protocol (UDP) and / or the internet protocol (IP) in the TCP / IP internet protocol suite. The networks 112 may include wired and / or wireless communications networks owned and / or operated by other service providers. For example, the networks 112 may include another CN connected to one or more RANs, which may employ the same RAT as the RAN 104 / 114 or a different RAT.

[0038] 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.

[0039] FIG. IB is a system diagram illustrating an example WTRU 102. As shown in FIG. IB, 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 elements / 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.

[0040] 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. IB 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, e.g., in an electronic package or chip.

[0041] 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 an 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 an embodiment, the transmit / receive element 122 may be configured to transmit and / or receive both RF and light signals. It will be appreciated that the transmit / receive element 122 may be configured to transmit and / or receive any combination of wireless signals.

[0042] Although the transmit / receive element 122 is depicted in FIG. IB as a single element, the WTRU 102 may include any number of transmit / receive elements 122. For example, the WTRU 102 may employ MIMO technology. Thus, in an 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.

[0043] 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.

[0044] 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 storedata 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), readonly 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).

[0045] 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.

[0046] 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.

[0047] The processor 118 may further be coupled to other elements / peripherals 138, which may include one or more software and / or hardware modules / units that provide additional features, functionality and / or wired or wireless connectivity. For example, the elements / peripherals 138 may include an accelerometer, an e-compass, a satellite transceiver, a digital camera (e.g., 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 elements / 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.

[0048] 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 uplink (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 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 WTRU 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 uplink (e.g., for transmission) or the downlink (e.g., for reception)).

[0049] 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, and 102c over the air interface 116. The RAN 104 may also be in communication with the CN 106.

[0050] 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 an embodiment, the eNode-Bs 160a, 160b, 160c may implement MIMO technology. Thus, the eNode-B 160a, for example, may use multiple antennas to transmit wireless signals to, and receive wireless signals from, the WTRU 102a.

[0051] Each of the eNode-Bs 160a, 160b, and 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 uplink (UL) and / or downlink (DL), and the like. As shown in FIG. 1C, the eNode-Bs 160a, 160b, 160c may communicate with one another over an X2 interface.

[0052] 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 (PGW) 166. While each of the foregoing elements are depicted as part of the CN 106, it will be appreciated that any one of these elements may be owned and / or operated by an entity other than the CN operator.

[0053] The MME 162 may be connected to each of the eNode-Bs 160a, 160b, and 160c in the RAN 104 via an SI 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 forswitching between the RAN 104 and other RANs (not shown) that employ other radio technologies, such as GSM and / or WCDMA.

[0054] The SGW 164 may be connected to each of the eNode-Bs 160a, 160b, 160c in the RAN 104 via the SI 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.

[0055] 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.

[0056] The CN 106 may facilitate communications with other networks. For example, the CN 106 may provide the WTRUs 102a, 102b, 102c with access to circuit-switched networks, such as the PSTN 108, to facilitate communications between the WTRUs 102a, 102b, 102c and traditional land-line communications devices. For example, the CN 106 may include, or may communicate with, an IP gateway (e.g., an IP multimedia subsystem (IMS) server) that serves as an interface between the CN 106 and the PSTN 108. In addition, the CN 106 may provide the WTRUs 102a, 102b, 102c with access to the other networks 112, which may include other wired and / or wireless networks that are owned and / or operated by other service providers.

[0057] 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.

[0058] In representative embodiments, the other network 112 may be a WLAN.

[0059] 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 into 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. l ie DLS or an802.1 Iz 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.

[0060] When using the 802.1 lac 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.

[0061] 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 nonadj acent 20 MHz channel to form a 40 MHz wide channel.

[0062] Very high throughput (VHT) STAs may support 20 MHz, 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 a medium access control (MAC) layer, entity, etc.

[0063] Sub 1 GHz modes of operation are supported by 802.1 laf and 802.11 ah. The channel operating bandwidths, and carriers, are reduced in 802.1 laf and 802.1 lah relative to those used in802.1 In, and 802.1 lac. 802.1 laf supports 5 MHz, 10 MHz and 20 MHz bandwidths in the TV white space (TVWS) spectrum, and 802.1 lah supports 1 MHz, 2 MHz, 4 MHz, 8 MHz, and 16 MHz bandwidths using non-TVWS spectrum. According to a representative embodiment,802.1 lah may support meter type control / machine-type communications (MTC), such as MTCdevices 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).

[0064] WLAN systems, which may support multiple channels, and channel bandwidths, such as 802.1 In, 802.1 lac, 802.11af, and 802.1 lah, 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 802.1 lah, 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.

[0065] In the United States, the available frequency bands, which may be used by 802.1 lah, 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 802.1 lah is 6 MHz to 26 MHz depending on the country code.

[0066] FIG. ID 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.

[0067] 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 an embodiment, the gNBs 180a, 180b, 180c may implement MIMO technology. For example, gNBs 180a, 180b may utilize beamforming to transmit signals to and / or receive signals from the WTRUs 102a, 102b, 102c. 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 multiplecomponent 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).

[0068] The WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c using transmissions associated with a scalable numerology. For example, OFDM symbol spacing and / or OFDM subcarrier spacing may vary for different transmissions, different cells, and / or different portions of the wireless transmission spectrum. The WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c using subframe or transmission time intervals (TTIs) of various or scalable lengths (e.g., including a varying number of OFDM symbols and / or lasting varying lengths of absolute time).

[0069] 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.

[0070] 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 functions (UPFs) 184a, 184b, routing of control plane information towards access and mobility management functions (AMFs) 182a, 182b, and the like. As shown in FIG. ID, the gNBs 180a, 180b, 180c may communicate with one another over an Xn interface.

[0071] The CN 115 shown in FIG. ID may include at least one AMF 182a, 182b, at least one UPF 184a, 184b, at least one session management function (SMF) 183a, 183b, and at least one 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.

[0072] 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 protocol data unit (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, e.g., 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 MTC access, and / or the like. The AMF 162 may provide a control plane function for switching between the RAN 113 and other RANs (not shown) that employ other radio technologies, such as LTE, LTE-A, LTE-A Pro, and / or non-3GPP access technologies such as WiFi.

[0073] The SMF 183a, 183b may be connected to an AMF 182a, 182b in the CN 115 via an N11 interface. The SMF 183a, 183b may also be connected to a UPF 184a, 184b in the CN 115 via an N4 interface. The SMF 183a, 183b may select and control the UPF 184a, 184b and configure the routing of traffic through the UPF 184a, 184b. The SMF 183a, 183b may perform other functions, such as managing and allocating UE 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.

[0074] 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, e.g., 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 multihomed PDU sessions, handling user plane QoS, buffering downlink packets, providing mobility anchoring, and the like.

[0075] 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 an 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.

[0076] In view of FIGs. 1 A-1D, and the corresponding description of FIGs. 1 A-1D, one or more, or all, of the functions described herein with regard to any of: WTRUs 102a-d, base stations 114a- b, eNode-Bs 160a-c, MME 162, SGW 164, PGW 166, gNBs 180a-c, AMFs 182a-b, UPFs 184a- b, SMFs 183a-b, DNs 185a-b, and / or any other element(s) / device(s) described herein, may be performed by one or more emulation elements / 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.

[0077] The emulation devices may be designed to implement one or more tests of other devices in a lab environment and / or in an operator network environment. For example, the one or more emulation devices may perform the one or more, or all, functions while being fully or partially implemented and / or deployed as part of a wired and / or wireless communication network in order to test other devices within the communication network. The one or more emulation devices may perform the one or more, or all, functions while being temporarily implemented / deployed as part of a wired and / or wireless communication network. The emulation device may be directly coupled to another device for purposes of testing and / or may performing testing using over-the-air wireless communications.

[0078] 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.

[0079] Provided below are acronyms / abbreviations for terms and phrases commonly used in this application:

[0080] Af Sub-carrier spacing

[0081] gNB NR NodeB

[0082] ACK Acknowledgement

[0083] Al Artificial Intelligence

[0084] AOA Angle of Arrival

[0085] AOD Angle of Departure

[0086] BLER Block Error Rate

[0087] BRS Beam Reference Signal

[0088] CA Carrier Aggregation

[0089] CB Contention-Based (e.g., access, channel, resource)

[0090] CG Cell Group

[0091] CoMP Coordinated Multi-Point transmission / reception

[0092] CP Cyclic Prefix

[0093] CP-OFDM Conventional OFDM (relying on cyclic prefix)

[0094] CQI Channel Quality Indicator

[0095] CN Core Network (e.g., LTE packet core or NR core)

[0096] CRC Cyclic Redundancy Check

[0097] CSI Channel State Information

[0098] CU Central Unit

[0099] D2D Device to Device transmissions (e.g., LTE Sidelink)

[0100] DC Dual Connectivity

[0101] DCI Downlink Control Information

[0102] DL Downlink

[0103] DM-RS Demodulation Reference Signal

[0104] DRB Data Radio Bearer

[0105] DU Distributed Unit

[0106] EN-DC E-UTRA - NR Dual Connectivity

[0107] EPC Evolved Packet Core

[0108] FDD Frequency Division Duplexing

[0109] FDM Frequency Division Multiplexing

[0110] IP Internet Protocol

[0111] LCH Logical Channel

[0112] LCP Logical Channel Prioritization

[0113] LOS Line of Sight

[0114] LTE Long Term Evolution e.g., from 3 GPP LTE R8 and up

[0115] MAC Medium Access Control

[0116] NACK Negative ACK

[0117] MBS Multicast Broadcast System

[0118] MBMS Multimedia Broadcast Multicast System

[0119] MCG Master Cell Group

[0120] MCS Modulation and Coding Scheme

[0121] MIMO Multiple Input Multiple Output

[0122] ML Machine Learning

[0123] MTC Machine-Type Communications

[0124] MR-DC Multi-RAT Dual Connectivity

[0125] NAS Non-Access Stratum

[0126] NE-DC NR-RAN - E-UTRA Dual Connectivity

[0127] NLOS Non-Line of Sight

[0128] NR New Radio

[0129] NR-DC Dual Connectivity with

[0130] OFDM Orthogonal Frequency-Division Multiplexing

[0131] OOB Out-Of-Band (emissions)

[0132] Pcmax Total available WTRU (e.g., UE) power in a given transmission interval

[0133] Pcell Primary cell of Master Cell Group

[0134] PCG Primary Cell Group

[0135] PDU Protocol Data Unit

[0136] PER Packet Error Rate

[0137] PHY Physical Layer

[0138] PLMN Public Land Mobile Network

[0139] PLR Packet Loss Rate

[0140] PRACH Physical Random Access Channel

[0141] PRB Physical Resource Block

[0142] PScell Primary cell of a Secondary cell group

[0143] PSS Primary Synchronization Signal

[0144] QoS Quality of Service (from the physical layer perspective)

[0145] RAB Radio Access Bearer

[0146] RAN PA Radio Access Network Paging Area

[0147] RACH Random Access Channel (or procedure)

[0148] RAR Random Access Response

[0149] RAT Radio Access Technology

[0150] RCU Radio access network Central Unit

[0151] RF Radio Front end

[0152] RLF Radio Link Failure

[0153] RLM Radio Link Monitoring

[0154] RNTI Radio Network Identifier

[0155] RRC Radio Resource Control

[0156] RRM Radio Resource Management

[0157] RS Reference Signal

[0158] RTT Round-Trip Time

[0159] SCG Secondary Cell Group

[0160] SCS Sub Carrier Spacing

[0161] SCMA Single Carrier Multiple Access

[0162] SDU Service Data Unit

[0163] SOM Spectrum Operation Mode

[0164] SpCell Primary cell of a master or secondary cell group.

[0165] SS Synchronization Signal

[0166] SSS Secondary Synchronization Signal

[0167] SRB Signaling Radio Bearer

[0168] SWG Switching Gap (in a self-contained subframe)

[0169] TB Transport Block

[0170] TBS Transport Block Size

[0171] TDD Time-Division Duplexing

[0172] TDM Time-Division Multiplexing

[0173] TI Time Interval (in integer multiple of one or more symbols)

[0174] TOA Time of Arrival

[0175] TRP Transmission / Reception Point

[0176] TRPG Transmission / Reception Point Group

[0177] TRx Transceiver

[0178] TTI Transmission Time Interval (in integer multiple of one or more symbols)

[0179] UL Uplink

[0180] URC Ultra-Reliable Communications

[0181] URLLC Ultra-Reliable and Low Latency Communications

[0182] V2X Vehicular communications

[0183] WLAN Wireless Local Area Networks and related technologies (IEEE 8O2.xx domain)

[0184] The following description is for exemplary purposes and does not intent to limit in any way the applicability of the methods described herein to any wireless technology, any communication technology and / or to other technology, when applicable. The term network in this disclosure may refer to one or more gNBs which in turn may be associated with one or more Transmission / Reception Points (TRPs), or to any other node in the radio access network.

[0185] Artificial intelligence may be broadly defined as the behavior exhibited by machines that mimics cognitive functions to sense, reason, adapt and act.

[0186] Machine learning may refer to the type of algorithms that solve a problem based on learning through experience ("data"), without explicitly being programmed ("configuring set of rules"). Machine learning can be considered as a subset of Al. Different machine learning paradigms may be envisioned based on the nature of data or feedback available to the learning algorithm. For example, a supervised learning approach may involve learning a function that maps input to an output based on labeled training example, wherein each training example may be a pair comprising (e.g., consisting of) input and the corresponding output. For example, unsupervised learning approach may involve detecting patterns in the data with no pre-existing labels. For example, reinforcement learning approach may involve performing sequence of actions in an environment to maximize the cumulative reward. In some solutions, it is possible to apply machine learning algorithms using a combination or interpolation of the above-mentioned approaches. For example, a semi-supervised learning approach may use a combination of a small amount of labeled data with a large amount of unlabeled data during training. In this regard, semi-supervised learning falls between unsupervised learning (with no labeled training data) and supervised learning (with only labeled training data).

[0187] Deep learning may refer to a class of machine learning algorithms that employ artificial neural networks (specifically DNNs), which were loosely inspired from biological systems. The Deep Neural Networks (DNNs) are a special class of machine learning models inspired from the human brain wherein the input is linearly transformed and passed through non-linear activation function multiple times. DNNs typically comprises (e.g., consists of) multiple layers where each layer comprises (e.g., consists of) linear transformation and a given non-linear activation function. The DNNs can be trained using the training data via back-propagation algorithm. Recently, DNNshave shown state-of-the-art performance in variety of domains, e.g., speech, vision, natural language etc., and for various machine learning settings, e.g., supervised, un-supervised, and semisupervised. An Al component may refer to realization of behaviors and / or conformance to requirements by learning based on data, without explicit configuration of sequence of steps of actions. Such Al component may enable learning complex behaviors which might be difficult to specify and / or implement when using legacy methods. An Al component can implement an Al model or a set of Al models.

[0188] Online learning may refer to a method of machine learning where data becomes available in a sequential manner, or generated as a function of time, and is used to update or train a model. Online learning allows a model to dynamically adapt to new patterns in the data when it appears over time.

[0189] In Federated Learning (FL), the central Al server trains a global model by combining local models trained by each participant (UE) based on a model averaging technique. The WTRU (e.g., UE) performs its local model training within each training cycle based on a model downloaded from the centralized Al server using local data. Once the local model training is completed, the WTRU (e.g., UE) delivers the training results (e.g., gradient for the Deep Neural Network - DNN) to the centralized Al server via 5G UL channels. Next, the centralized Al server aggregates the gradients (i.e., model parameters) from the WTRUs (e.g., UEs), and updates the global model. Then, the next training cycle begins where the Al server would distribute the updated global model to WTRUs (e.g., UEs) via 5G DL channels. FIG. 2 shows a high-level interaction between participants and the central Al server over the 5G system.

[0190] The FL training over wireless communication is slightly different than the FL training in data centers, where participants (UEs) may have highly variable conditions in terms of available computational and network resources. Additionally, the WTRUs (e.g., UEs) may not be homogeneous, so they may have different capabilities in terms of their computing and network- related resources or even what ML frameworks they support. Thus, it may not be efficient for a centralized Al server to include all the WTRUs (e.g., UEs) (participants) in a training session, so some sort of member selection mechanism is needed before each training cycle begins. Note should be taken that if the conditions (e.g., device's computation resource and / or wireless channel condition) are not changed, the WTRU (e.g., UE) re-selection and training re-configurations might not be needed for each training cycle. However, it is crucial to re-select different WTRUs (e.g., UEs) over time to achieve global training with diverse datasets.

[0191] FIG. 3 highlights an FL scenario where a set of participants (i.e., Devices A-E) is involved in a distributed training session. However, all end devices (WTRUs (e.g., UEs)) are not involvedin all training cycles. In some training cycles, some WTRUs (e.g., UEs) are inactive, while in other sessions, they are busy training a local model. For example, Device A is initially (during the N-th cycle) engaged in a training session, but after it reports its training resources, the centralized FL server does not select Device A for the next cycle (i.e., N+l-th cycle); instead, it selected Device B which is inactive during the N-th cycle. Note should be taken that there are a lot of times that the WTRU (e.g., UE) or the FL server may be inactive between and during the training sessions.

[0192] Each FL training cycle can be categorized into three operating stages. In the first operating stage (yellow-colored box in FIG. 3), the FL training server selects a set of training devices (WTRUs (e.g., UEs)). During this stage, the training devices should first express their training resources to the FL training server. Once the FL server collects all information from training devices, it can select some of them, and then it enters the next operating stage (i.e., model distribution and training configuration, i.e., the green-colored box in FIG. 3, where the FL server distributes a global trained model and related configurations to all selected devices (WTRUs (e.g., UEs)). Upon the new model's arrival and configuration, WTRUs (e.g., UEs), start local model training, perhaps at a different point in time. Once the local training is completed at a UE, it is delivered to the FL server, where the third operating stage begins (i.e., the pink-colored box in FIG. 3. During this stage, the FL server aggregates all training results to form a global model and repeats the whole training workflow again.

[0193] 3GPP has started specifying mechanisms and frameworks for using AI / ML based approaches both the air interface level (e.g., CSLfeedback enhancements (e.g., CSI compression and CSI prediction), beam management, and positioning accuracy enhancement) and network level (e.g., network energy saving, load balancing and mobility).

[0194] AI / ML operations are based on models (e.g., neural networks) that have been trained using a substantial amount of data under different scenarios / conditions. For a given function, there could be several models (e.g., each trained / suitable for different network / UE condi ti ons / characteri sti cs) .

[0195] Once a model is well trained, it can be deployed (e.g., in a test environment, e.g., test network) for performance testing. Model monitoring may need to be performed even after deployment in a real network, as the current network / WTRU (e.g., UE) conditions can become different from the scenarios / conditions in which the model was trained / tested. If the model monitoring is shown to provide undesirable WTRU (e.g., UE) / network performance, a decision could be made to switch to another model or stop using AI / ML based operations for the concerned function, etc., The performance monitoring can also be used to determine whether a model needs to be retrained with new sets of data.

[0196] Model training and monitoring can be performed at the WTRU (e.g., UE), at the network, or in collaboration between the two. Model training and monitoring can be performed offline or online.

[0197] As introduced in Rel-18 study item on AI / ML for NR air interface, there are two main categories of AI / ML models based on the inference operation as one-sided model (i.e., the inference is done at either the network (NW) side or the WTRU (e.g., UE) side) and two-sided model (i.e., the inference is done sequentially between the WTRU (e.g., UE) and the NW or vice- versa).

[0198] There are several challenges associated with supporting Al in a communication system, or within a communication protocol stack.

[0199] Methods and apparatus allowing less model management complexity with nested AI / ML model architecture than with multiple AI / ML models may be provided.

[0200] Managing a multitude of AI / ML models for the same function to support different capabilities of a device / node requires more resources and adds complexity e.g., storage, model transfer, dataset transfer, and training, etc. Alternatively, enabling a single nested AI / ML model that can scale according to the capabilities of a device / node and / or dynamic conditions can alleviate these limitations.

[0201] Methods and apparatus allowing flexibility for the WTRU / NW to adapt the nested AI / ML model based on dynamic conditions may be provided.

[0202] There are several challenges for developing a single AI / ML model for Each (e.g., one or more) dynamic condition, e.g., link conditions (RSRP, RSRQ, SINR, etc.), channel types (UMi, UMa, indoor, outdoor, etc.), QoS requirements, WTRU (e.g., UE) RRC states, WTRU (e.g., UE) mobility, etc. First, the frequent AI / ML model transfer increases the AI / ML-related signaling overhead. Second, storing several AI / ML models leads to storage issues. Alternatively, a single nested AI / ML model architecture, through its flexibility, may remove the necessity of the transfer / storage of several AI / ML models for different dynamic conditions.

[0203] Methods and apparatus allowing larger training dataset size for nested AI / ML model representing all dynamic conditions and / or device / node capabilities may be provided.

[0204] Training a multitude of AI / ML models for the same function introduces the challenge of insufficient / small training dataset size. Even though the total dataset size (i.e., representing all device / node capabilities and dynamic conditions) is large, an AI / ML model targeting a particular case is trained via a comparatively smaller dataset size. On the other hand, a single nested AI / ML model architecture is composed of multiple sub-models, where Each (e.g., one or more) sub-modelis for a particular case. Hence, the nested AI / ML model architecture allows the training of submodels to benefit from joint utilization or reuse of the available datasets.

[0205] Methods and apparatus for distributed / federated learning of nested models may be provided.

[0206] The Nested AI / ML model architecture may enable the WTRUs (e.g., UEs) with diverse capabilities to collaboratively contribute to the training of an AI / ML model in the context of distributed / federated learning. Federated learning also has the advantage of alleviating the overhead of sending the local datasets to the NW while keeping the privacy and security of the data local to the UEs.

[0207] One or more examples described herein may be used separately or in combination for the WTRU to determine distributed / federated learning of nested models.

[0208] The WTRU (e.g., UE) may receive a request to provide its capabilities and / or dynamic condition to the NW.

[0209] The WTRU (e.g., UE) may indicate, for example in UE-NR-Capability, any of the following capabilities: (1) its compute capabilities, which indicate the processing capabilities or processing profile that may be expressed in terms of compute features or architectures, raw compute metrics associated with the compute architecture, etc., or (2) functions(s) identifier (s) or component(s) identifier(s) for which an Al component can be configured with (e.g., online) learning capabilities.

[0210] The WTRU (e.g., UE) may indicate one or more dynamic conditions that may affect the UE ability to provide compute or memory resources to ML task(s) or that may limit signaling required to support ML task(s) e.g., WTRU (e.g., UE) power saving state, radio resources, QoS, latency, available compute resources, etc.

[0211] The WTRU (e.g., UE) may be configured to perform (e.g., online) learning for a given Al component with a nested sub-model ID.

[0212] The WTRU (e.g., UE) may receive one or more of the following configurations: (1) one or more Al component identifier (e.g., Al component that may identify the function(s) being replaced by an AI / ML model); (2) one or more learning parameters (e.g., loss function, number of training iterations, loss threshold, learning method, etc.); or (3) one or more nested model attributes: (i) sub-model identifier; or (ii) sub-model attributes (e.g., input data size / type, output data size / type).

[0213] The WTRU (e.g., UE) may receive an AI / ML nested model for the configured Al component.

[0214] The WTRU (e.g., UE) may receive an indication to start the (e.g., online) learning for the configured Al component and nested model sub-model ID.

[0215] The WTRU (e.g., UE) may apply (e.g., online) learning procedures for the configured Al component with nested model sub-model ID using local data.

[0216] The WTRU (e.g., UE) may transmit the trained AI / ML nested model parameters and number of training iterations corresponding to the nested model sub-model ID.

[0217] The WTRU (e.g., UE) may receive an updated AI / ML nested model for the configured Al component.

[0218] Methods and apparatus for hierarchical federated learning of nested models may be provided.

[0219] The Nested AI / ML model architecture may enable the WTRUs (e.g., UEs) with diverse capabilities to collaboratively contribute to the training of an AI / ML model in the context of distributed / federated learning. Federated learning also has the advantage of alleviating the overhead of sending the local datasets to the NW while keeping the privacy and security of the data local to the WTRUs (e.g., UEs). Hierarchical federated learning (or decentralized federated learning) may enable grouping of WTRUs (e.g., UEs) with similar capabilities and / or dynamic conditions and uses Sidelink or UE-to-UE direct communication between WTRUs (e.g., UEs) to aggregate learned weights within Each (e.g., one or more) group or cluster of WTRUs (e.g., UEs). A WTRU (e.g., UE) may be configured as an aggregator WTRU (e.g., UE) within Each (e.g., one or more) group or cluster of WTRUs (e.g., UEs). The aggregator WTRU (e.g., UE) may transmit to the NW with the aggregated model parameters for a group or cluster of UEs.

[0220] One or more examples described herein may be used separately or in combination for the WTRU to determine hierarchical federated learning of nested models.

[0221] The WTRU (e.g., UE) may receive a request to provide its capabilities and / or dynamic condition to the NW. The WTRU (e.g., UE) may indicate, for example in UE-NR-Capability, its compute capabilities, functions(s) or component(s) for which an Al component can be configured with (e.g., online) learning capabilities, etc. The WTRU (e.g., UE) may indicate dynamic conditions e.g., WTRU (e.g., UE) power saving state, radio resources, QoS, latency, available compute resources, etc.

[0222] The WTRU (e.g., UE) may be configured to perform (e.g., online) learning for a given Al component with aggregator WTRU (e.g., UE) identity and / or identity of WTRUs (e.g., UEs) in the group or cluster of WTRUs (e.g., UEs). The Al component may identify the function(s) being replaced by the AI / ML nested model at the WTRU (e.g., UE). The aggregator WTRU (e.g., UE) identity may provide the WTRU (e.g., UE) with the WTRU (e.g., UE) identity to whichAI / ML nested model parameters shall be sent to i.e., the WTRU (e.g., UE) configured for doing local aggregation via Sidelink or WTRU-to-WTRU (e.g., UE-to-UE) direct communication within the configured group or cluster of UEs. If the WTRU (e.g., UE) is configured as the aggregator UE WTRU (e.g., UE) the WTRU (e.g., UE) may receive the identity of WTRUs (e.g., UEs) in the group or cluster of WTRUs (e.g., UEs).

[0223] The WTRU (e.g., UE) may receive an AI / ML nested model for the configured Al component with (e.g., online) learning.

[0224] If the WTRU (e.g., UE) is configured as the aggregator WTRU (e.g., UE): (1) the WTRU (e.g., UE) may transmit an AI / ML nested model to the WTRUs (e.g., UEs) in the configured group or cluster of WTRUs (e.g., UEs) via Sidelink or WTRU-to-WTRU direct communication; and / or the WTRU (e.g., UE) may transmit an indication to the WTRUs (e.g., UEs) in the configured group or cluster of WTRUs (e.g., UEs) via Sidelink or UE-to-UE direct communication to start the (e.g., online) learning of a configured Al component and nested model sub-model ID.

[0225] The WTRU (e.g., UE) may apply (e.g., online) learning procedures for the configured Al component using local data.

[0226] If the WTRU (e.g., UE) is configured as the aggregator WTRU (e.g., UE), the WTRU may perform any of the following actions: (1) the WTRU (e.g., UE) may receive via Sidelink or WTRU-to-WTRU direct communication, trained AI / ML nested model parameters with corresponding sub-model ID from other WTRUs (e.g., UEs) in the configured group or cluster of WTRUs (e.g., UEs); (2) the WTRU (e.g., UE) may determine the nested model sub-model ID to perform aggregation based on received trained nested model parameters and associated sub-model IDs; or (3) the WTRU (e.g., UE) may transmit locally aggregated nested model parameters to the NW.

[0227] The methods described herein are applicable, without limitation to, any communication link that include two (point-to-point) or more (point-to-multipoint) communication devices such as 3GPP LTE Uu, 3GPP NRUu, 3GPP Sidelink, IEEE 802.11 / 15 technologies including protocols for wireless air interfaces and device-to-device communications.

[0228] The invention may enable the WTRUs (e.g., UEs) with different static capabilities and / or various dynamic conditions to collaboratively contribute to the training of a configurable nested AI / ML model within the context of distributed / federated learning. Each (e.g., one or more) WTRU (e.g., UE) with its dataset may train an AI / ML model as a part of the nested AI / ML model, then the trained AI / ML models are aggregated according to the complete nested AI / ML model architecture. Hence, this invention may enable the utilization of various local datasets from dispersed WTRUs (e.g., UEs) in the training / development of a configurable nested AI / ML model.

[0229] A nested model architecture is proposed.

[0230] Methods are described to enable a flexible model architecture in a system where a communication protocol may be configured as a chain of functions. Possibly within the processing path of a layer, modeled as processing block (or plurality of processing blocks, cascaded or otherwise). Each (e.g., one or more) processing block, function or a combination of multiple functions may be configured as a rule-based function or as an Al component (e.g., an AI / ML model).

[0231] In some approaches, the term "rule-based" processing may refer to specified WTRU (e.g., UE) behavior and / or requirements explicitly defined in the form of procedural text, signaling syntax or the likes. Rule based processing may refer to any processing based on legacy algorithms that are essentially non-AI based.

[0232] Furthermore, in some solutions, each (e.g., one or more) function may be configured as a combination of a rule-based part and an AI / ML model part, where Each (e.g., one or more) part has a specific input and output. Possibly a WTRU (e.g., UE) and / or a NW may be configured via a nested AI / ML model architecture either for the same function or for the combination of multiple functions. Precisely, the nested AI / ML model architecture is a hierarchical AI / ML model that includes smaller models nested inside a larger model. Here, the model size (small vs. large) is defined in terms of input data size, number of layers and other attributes.

[0233] FIG. 4 illustrates an example of nested AI / ML model architecture with three sub-models named as Model A, Model B, and Model C, respectively. Here, Model A, which is the smallest model with the lowest input data size and / or the lowest number of layers, is nested in Model B. In addition, Model B, which is defined as the combination of Model A and Model B', is nested in Model C, which is the largest model with higher input data size or more layers, and it includes both sub-models (i.e., Model A and Model B). Hence, each (e.g., one or more) sub-model may be identified with a sub-model ID and its corresponding distinct attributes. In other words, the different variations of nested models can be envisaged with respect to the design attributes of Each (e.g., one or more) sub-model. These variations may include any of the following: (1) input data size / type; (2) output data size / type; (3) dataset size for training, test, and validation; (4) number of layers; (5) types of Each (e.g., one or more) layer (e.g., fully connected, convolutional, recurrent, pooling, etc.); (6) bit-depth (i.e., quantization level for model parameters); or (7) loss function.

[0234] The above attributes may be part of the configuration provided to the WTRU (e.g., UE) for the nested model. Other design attributes for the nested models may also be included (e.g., model pruning level, convolutional kernel dilation, stride length, or other applicable attributes). Each (e.g., one or more) sub-model in the nested architecture may have its own pre-processing andpost-processing routines given the scenario of different input and output data sizes / types, respectively.

[0235] Each (e.g., one or more) sub-model may be classified in terms of both performance and complexity (e.g., the computational complexity of model inference / training, the number of model parameters, the model size, the computational complexity for pre- and post-processing, etc.). Therefore, for a given single nested AI / ML model, the NW may request the WTRU (e.g., UE) to use either a smaller or larger sub-model based on static capabilities (e.g., number of antennas, compute capability, etc.) that may affect the UE ability to provide compute or memory resources to ML task(s) or that may limit signaling required to support ML task(s) and / or dynamic conditions (UE power saving state, radio resources, QoS, latency, available compute resources, link conditions, channel types, etc.).

[0236] Variations and Use Cases for Nested AI / ML Model Architectures is described hereafter.

[0237] A communication protocol may be configured as a chain of functions. The nested AI / ML model architecture may include multiple sub-models for the same function or for the combination of multiple functions. Here, Each (e.g., one or more) sub-model is identified with a set of its distinct design attributes, which may include the input / output data size / type, the dataset size, the number of layers, the type of Each (e.g., one or more) layer, the bit-depth (i.e., quantization level for model parameters) of Each (e.g., one or more) sub-model, the loss function, etc. As a design criterion, the sub-models inside the full nested model may share either the same or different set of attributes. Hence, in addition to the nested AI / ML model architecture example presented in FIG. 4, there are diverse variations of the nested AI / ML model architectures.

[0238] Examples of seven possible variations of nested AI / ML model architectures are illustrated in FIG. 5, FIG. 6, FIG. 7, and FIG. 8. There may be other variations beyond these examples. Here, each (e.g., one or more) variation includes three sub-models: Model A, Model B, and Model C. All variations are explained in detail as follows:

[0239] FIG. 5 illustrates variations of nested AI / ML model architectures based on input data size and layers.

[0240] Variation #1 : FIG. 5(a) illustrates a general concept for the nested AI / ML model architectures, where Model A is nested within Model B and Model B is nested within Model C. Hence, Model B is composed of both Model A and Model B'. Similarly, Model C is developed by combining Model A, Model B', and Model C. Here, the smallest sub-model (i.e., Model A) has the lowest input data size and includes the lowest number of layers compared to the larger models (i.e., Model B and Model C). In other words, the larger models have more layers and higher input data size for the purpose of improving the accuracy / performance and / or including additionalfunctionalities. Particularly, Model A (e.g., only) uses one set of inputs (i.e., Input Set #1), while Model C exploits all three sets of inputs (i.e., Input Set #1, Input Set #2, and Input Set #3). Regarding the model outputs, there might be several options. First, when Each (e.g., one or more) sub-model is trained / developed for the same function(s), then all models may have the same output data size. Second, the larger models may make additional predictions, which leads to larger output data size. Third, when a particular model (e.g., Model B) includes additional functionalities compared to a smaller model (e.g., Model A), then Each (e.g., one or more) model may have different output data size and / or type (e.g., the output size of the corresponding larger model may be either smaller or larger than that of smaller model). Regarding the computational complexity, in this variation, the smaller model has lower computational complexity since they are nested inside the larger models with additional layers.

[0241] Variation #2: FIG. 5(b) presents a nested AI / ML model architecture with disjoint submodels inside the full nested model. Particularly, both Model A and Model B are disjointly nested in Model C. Even though Model A and Model B take the inputs from different sets of inputs, they may have the same number / type of layers and / or the same input / output data size / types. On the other hand, Model C has more layers to exploit both sets of inputs (i.e., Input Set #1 and Input Set #2). In this variation, Model A and Model B may have either the same or similar computational complexity, while Model C has higher computational complexity.

[0242] Variation #3: FIG. 5(c) shows a nested AI / ML model architecture, where there are two sets of inputs. Here, Model A is nested in Model B as in the Variation #1. However, both Model A and Model B have the same input data size / type since they employ the same set of inputs (i.e., Input Set #1) at their input layer. In comparison to Model A, Model B employs additional layers as indicated by Model B'. Here, Model B may target to improve the accuracy / performance and / or make additional predictions with the expense of an increased computational complexity. Even though Model A, as a sub-model nested in Model B, utilizes the Input Set #1 at its input layer, Model B may also reuse the same Input Set #1 at the input layer of Model B'. Like Variation #1 and Variation #2, Model C with larger input data size (i.e., utilizing both Input Set #1 and Input Set #2) and more layers aims to further enhance the accuracy / performance.

[0243] Variation #4: FIG. 5(d) demonstrates a nested AI / ML model architecture, where all three sub-models have the same input data size / type. Therefore, unlike the previous variations, (e.g., only) Input Set #1 is utilized for Each (e.g., one or more) sub-model. On the other hand, the larger sub-models, with more layers, may target to improve the accuracy / performance and / or to include additional functionalities and / or to make additional predictions. This implies that the output layer size of Each (e.g., one or more) sub-model may vary.

[0244] Variation #5 : FIG. 6 presents an example for the nested AI / ML model architecture having different bit-depths for Each (e.g., one or more) sub-model. For example, there may be Q versions for Model A, where Each (e.g., one or more) version represents a specific quantization level (i.e., Model A£denotes the ith version of Model A with i G {1,2, ... , Q}). Similarly, Model B and Model C may also have Q versions. Please note that Each (e.g., one or more) sub-model may have a different number of versions and / or different quantization levels. Based on the accuracy / performance requirement and / or available compute capability, either NW or WTRU (e.g., UE) may select a specific quantization level for Each (e.g., one or more) model. Afterwards, the nested AI / ML model may be constructed as a combination of multiple sub-models with different quantization levels. For instance, the nested AI / ML model shown in FIG. 5 is created from Model A£, Model Bj, and Model C{ with i,j, k G {1,2, ... , Q}. Then, Model B is shaped from Model A£and Model Bj. Therefore, there may be separate quantization levels for smaller sub-models nested inside larger sub-models. It is important to mention that when the bit-depth is considered, in addition to Variation #5, the nested AI / ML model may also follow other architectures, i.e., the variations presented in FIG. 5.

[0245] Variation #6: FIG. 7 shows an example of two-sided AI / ML model architecture. According to the autoencoders, each (e.g., one or more) sub-model may comprise (e.g., consist of) of two parts (i.e., encoder and decoder). For instance, Model Aeand Model Ad denote encoder and decoder of two-sided Model A, respectively. This variation may be utilized for the use cases requiring two-sided AI / ML models, which require jointly running the inference at two different entities. Then, the encoder nested AI / ML model may be deployed at the NW side, while the decoder nested AI / ML model may be deployed to the WTRU (e.g., UE) side, or vice versa. When one side utilizes a particular sub-model for the encoder, the other side selects the corresponding sub-model for the decoder considering the joint inference. In addition to Variation #6, the two- sided nested AI / ML model architecture may also follow other variations on Each (e.g., one or more) side, e.g., the variations shown in FIG. 5.

[0246] Variation #7: FIG. 8 presents an example of nested AI / ML model architectures with local AI / ML models. We here consider there are K WTRUs (e.g., UEs), and Each (e.g., one or more) may use different sub-models. For instance, both UEi and UE2 utilize Model A, while UE / employs Model C according to their static capabilities and / or dynamic conditions. Also, each (e.g., one or more) WTRU (e.g., UE) may employ and train their own local AI / ML models (e.g., Model ai, Model a2, ..., Model c / ; in FIG. 8), which is connected to a corresponding sub-model (e.g., Model A, Model B, or Model C). These local AI / ML models may not be shared among WTRUs (e.g., UEs). For example, in FIG. 8, although UEI and UE2 utilize the same Model A, the inputdata of UE1 first enters to the local Model al. On the other hand, UE2 employs its local model as Model a2 prior using the common Model A. Here, the local models (i.e., Model al and Model a2) may have different attributes (e.g., input data size / type, number of layers, types of Each (e.g., one or more) layer, etc.), while their output data size / type should be the same. Since the outputs of these local models enter to the same common sub-model (i.e., Model A). Here, the training of the nested AI / ML model may be done through distributed / federated learning. However, the training of local AI / ML models (i.e., al, a2, . . ., cK) may be trained locally at the corresponding WTRUs (e.g., UEs). Furthermore, for the sake of generality, a local AI / ML model may be located at the output (or inside) of the corresponding sub-model from the nested AI / ML model architecture. The WTRU (e.g., UE) may need to inform the NW for the utilization and location of local AI / ML models in connection with the sub-models.

[0247] The proposed nested AI / ML model architecture can be utilized to support a wide range of use cases in a communication system or within a communication protocol stack. In general, the nested AI / ML models may be implemented for at least two approaches: (i) AI / ML-based schemes to replace the existing rule-based function(s), (ii) AI / ML-assisted schemes to enhance the existing rule-based function(s).

[0248] Here, six exemplary use-cases employing the proposed nested AI / ML model architectures are provided along with the captured use cases in Rel-18 and beyond:

[0249] Exemplary Use Case #1 : Time-domain CSI prediction is one sub-use case within the AI / ML for CSI feedback enhancement. Here, the input / output size of the AI / ML model depends mainly on the channel matrix size, which is a function of the number of transmit antennas (IVr), the number of receive antennas (1VR), and the number of sub-bands (Nsub). Traditionally, as NTand / or NRand / or Nsubvaries, it may be necessary to develop / train a multitude of AI / ML models (e.g., WTRUs with different number of antennas, allocation of different number of sub-bands, etc.). Alternatively, a single AI / ML model may be designed via the nested AI / ML model architecture with Variation #1 illustrated in FIG. 5(a). The larger models (e.g., Model C) may support larger input / output data size, implying a higher number of transmit / receive antennas and / or a higher number of sub-bands. Analogously, the smaller models (e.g., Model A) may be utilized, when there are lower number of transmit / receive antennas. Furthermore, when the number of subbands changes, a WTRU (e.g., UE) having the nested AI / ML model may switch between the submodels for the CSI prediction. Hence, it may both remove the requirement of storing multiple AI / ML models for the same function(s) and reduce the AI / ML-related signaling overhead for the AI / ML model transfer.

[0250] Exemplary Use Case #2: Spatial-Frequency Domain CSI compression is one sub-use case within the AI / ML for CSI feedback enhancement. It aims to reduce the feedback overhead size while sending CSI (e.g., raw channel matrix or eigenvectors). Since the CSI compression includes both encoder and decoder for the compression and reconstruction tasks, respectively, it is necessary to employ a two-sided nested AI / ML model architecture. Hence, Variation #6 shown in FIG. 7 may be utilized for the CSI compression. As in Exemplary Use Case #1, the input / output size of the AI / ML model depends on the channel matrix size (i.e., the number of transmit antennas (IVr), the number of receive antennas (1VR), and the number of sub-bands (Nsub)). Then, encoder and decoder models may be utilized at the WTRU (e.g., UE) size and the NW side, respectively. For instance, Model Ce may be deployed to the WTRU (e.g., UE) for the CSI compression, while the NW may use Model Cd for the CSI reconstruction. Afterwards, both WTRU (e.g., UE) and NW may select another sub-model together (e.g., Model Be and Model Bd) as the channel size changes (e.g., a lower number of sub-bands is allocated).

[0251] Exemplary Use Case #3: Spatial domain beam prediction is one sub-use case of AI / ML for beam management, which aims to predict Top-K beams based on Ll-RSRP feedback. Here, K is an important design attribute for the AI / ML models since the output data size may vary for different values of K (i.e., Top-1, with only 1 output, while Top-5 with 5 outputs). In contrast, the input data size depends on Ll-RSRP. Afterwards, the nested AI / ML model architecture with Variation #4 shown in FIG. 5(d) may be developed / trained for the spatial domain beam prediction. Thus, the same set of inputs (i.e., Ll-RSRP) may be utilized for Top-K beam prediction, where a smaller model with less layers (e.g., Model A) may predict Top-5 beams. With further refinement via more layers, a larger model (e.g., Model C) may predict Top-1 beam.

[0252] Exemplary Use Case #4: Direct AI / ML positioning is one sub-use case of AI / ML for positioning accuracy enhancements, where the AI / ML models target to predict two-dimensional horizontal position information. Regarding the model inputs, there are two options as channel impulse response (CIR) and power delay profile (PDP). Furthermore, the input data size is defined as NTRPX NportX Nt, where NTRPis the number of TRPs, Nportis the number of transmit / receive antenna port pairs, and Ntis the number of time domain samples. The nested AI / ML model architecture with Variation #2 presented in FIG. 5(b) may be used for the direct AI / ML positioning, where Input Set #1 and Input Set #2 may be CIR and PDP, respectively. Such architecture employs two disjoint sub-models with different sets of inputs. Particularly, Model A (e.g., only) utilizes CIR to predict the position information, whereas Model B (e.g., only) uses PDP for the same purpose. On the other hand, Model C as a larger AI / ML model exploits both CIR and PDP.

[0253] Exemplary Use Case #5: AI / ML-assisted positioning is another sub-use case of AI / ML for positioning accuracy enhancements, where the AI / ML models predict the intermediate performance metrics for the positioning (i.e., LoS / NLoS classification, ToA, RSTD, AoA / AoD, etc.). The nested AI / ML model architecture with Variation #3 shown in FIG. 5(c) may be utilized for the positioning accuracy enhancements, where Input Set #1 and Input Set #2 may be CIR and PDP, respectively. Here, Model A nested in Model B may predict the intermediate positioning performance metrics for the AI / ML-assisted positioning. Later, for the direct AI / ML positioning Model B may be designed by using the outputs of Model A and the same set of inputs (e.g., CIR). As in Model B, Model C may be designed for the direct AI / ML positioning, however, it utilizes use both CIR and PDP.

[0254] Exemplary Use Case #6: Lastly, the constellation learning may be considered as a use case for the nested AI / ML model architectures. The learned constellation phenomenon implies that that the constellation designed by AI / ML techniques may improve the error performance compared to the conventional constellations (e.g., QAM, PSK). According to the constellation size and the required bit rate, the various sub-models inside the nested AI / ML model architectures may be utilized for the constellation learning.

[0255] Methods and apparatus to introduce and / or control the use of nested models within the protocol may be proposed.

[0256] A protocol layer may be defined using one or more processing blocks. Each (e.g., one or more) processing block may have well defined / specific inputs and outputs. Herein the processing blocks can be either implemented as rule-based steps or using an Al component. In some solutions, the Al component may be implemented as a nested model. The nested model may be semi- statically or dynamically configured to enable scalable model operations e.g., training, transfer, or inference. Possibly, the selection of nested model attributes or sub-model may be impacted by one or more of the following: NW configuration, WTRU (e.g., UE), implementation or capabilities, application configuration or default / reference Al nested model configuration. Further the selection of nested model attributes or sub-model may be impacted by one or more of the following dynamic conditions: WTRU (e.g., UE) power saving state, radio resources, QoS, latency, available compute resources. The nested model may also be configurable to achieve different levels of performance e.g., configurable processing complexity / accuracy / power / granularity etc.

[0257] The nested model may also be configured by the NW to enable nested model attribute selection or sub-model selection at the WTRU (e.g., UE) based on dynamic conditions e.g., WTRU (e.g., UE), power saving state, radio resources, QoS, latency. Further the nested model may beconfigured by the NW for semi-static operations where the NW control the nested model attributes for a given Al component.

[0258] Methods and apparatus for signaling aspects associated with configuration and management of nested models may be proposed.

[0259] In a solution, a WTRU (e.g., UE) may receive a nested model configuration via unicast RRC signaling, possibly as a part of RRC reconfiguration procedure. For example, a WTRU (e.g., UE) may indicate its capability in terms of storage availability, support of various Al components, parametrizations, processing latency, compute capability, etc. Possibly as part of an RRC connection request or any other RRC message. The WTRU (e.g., UE) may receive in a RRC connection setup or RRC reconfiguration, the configuration associated with one or more nested models. For example, a WTRU (e.g., UE) may be configured to activate a nested model configuration when the RRC reconfiguration is applied. In another example, the WTRU (e.g., UE) may be configured to activate a set of the nested model attributes and / or sub-model upon an activation command via MAC CE (Control Element), DCI, or the likes. For example, the set of nested model attributes to activate may be referred to by a sub-model identifier.

[0260] In a solution, the nested model configuration may include attributes like, but not limited to, configured set of INPUTs and OUTPUTS of variable dimensions, a set of variable number of layers supported by the sub-models as part of the nested model. In addition, the configuration may contain a set of rules for pre-processing of INPUTs or post-processing of OUTPUTS. Other rules may be used for sub-model activation, selection, or deactivation. In a solution, the nested model attributes or sub-models may provide different performance profile and / or thresholds to enable the WTRU (e.g., UE) to adapt to dynamic conditions such as WTRU (e.g., UE) power saving state, radio resources, QoS, latency, available compute resources, available data, etc. with using the nested model.

[0261] In another solution, RRC signaling, MAC CE or SIB based signaling may be used to configure a group of WTRUs (e.g., UEs) sharing similar capabilities and / or dynamic conditions with the same nested model or sub-models. These WTRUs (e.g., UEs) may be configured to perform on-line learning of the nested model or sub-models. The WTRU (e.g., UE) may select a sub-model based on its dynamic conditions and / or available data for the model learning operation. The WTRU (e.g., UE) may update the NW or another WTRU (e.g., UE) with learned weights, gradients, number of training iterations, dataset size, trained sub-model ID, etc.

[0262] A WTRU (e.g., UE) performing a training iteration is where the device updates some or all the weights of a local model as part of training.

[0263] A weight update transmission is where a device may send updates for some or all of the weights of a local model which may combine multiple training iterations.

[0264] A model update is where a node or a device may receive one or more weight update transmissions, and / or performs one or more training iterations, and thru aggregation of updated weights, a global model is updated.

[0265] Methods and apparatus for distributed / Federated Learning of Nested Models may be proposed.

[0266] The NW may define, train, and deploy a nested AI / ML model architecture in a distributed manner using the WTRUs (e.g., UEs) compute capabilities and their own dataset. First, the NW may send the nested model configuration to the WTRUs (e.g., UEs) based on their compute capabilities and / or dynamic conditions. Second, each (e.g., one or more) WTRU (e.g., UE) may train the associated nested model parameters using its own dataset. Finally, each (e.g., one or more) WTRU (e.g., UE) may send the trained nested model parameters to the NW, which in turns aggregates them to the global AI / ML nested model.

[0267] A WTRU (e.g., UE) may be configured with a nested AI / ML model for an Al component. In a solution, the WTRU (e.g., UE) may be configured with a sub-model ID which may identify a part of the nested AI / ML model that is to be trained by the WTRU (e.g., UE) with using local data. In a solution, the configuration may include, but not limited to, learning parameters with number of training iterations, learning method, a loss function, nested model attributes with sub-model ID, performance thresholds.

[0268] Additionally, in a solution, the WTRU (e.g., UE) may be configured to select a sub-model ID based on performance thresholds. This configuration may enable the WTRU (e.g., UE) to perform training and inference with consideration for the WTRU (e.g., UE) static capabilities and / or dynamic conditions with using the same nested model.

[0269] In a solution, a WTRU (e.g., UE) with a configured nested model in the context of distributed / federated learning, leverages the nested model architecture to enable the training of sub-models based on WTRU (e.g., UE) static capabilities and / or dynamic conditions.

[0270] Specifically, at Each (e.g., one or more) iteration of the distributed / federated learning, the operation of nested models is depicted in FIG. 9.

[0271] The NW may request the capabilities of all WTRUs (e.g., UEs) (i.e., static capabilities and dynamic conditions) (S910).

[0272] Each (e.g., one or more) WTRU (e.g., UE) may send its static capabilities and dynamic conditions to NW (S911).

[0273] The NW may cluster the WTRUs (e.g., UEs) into groups based on their static capabilities and / or dynamic conditions (S912).

[0274] The NW may determine the nested AI / ML model architecture for a given Al component (S912).

[0275] The NW may map the WTRU (e.g., UE) groups to sub-models in the selected nested AI / ML model (S912).

[0276] The NW may send the nested AI / ML model and the associated attributes, including but not limited to, sub-model ID, to the corresponding WTRUs (e.g., UEs) (S913).

[0277] The NW may request the WTRU (e.g., UE) to start training of nested model sub-model ID with learning parameters like number of training iterations (S914).

[0278] The WTRU (e.g., UE) may acknowledge the training request (S915).

[0279] Each (e.g., one or more) WTRU (e.g., UE) may train its associated nested model parameters using its own data set (S916).

[0280] Each (e.g., one or more) WTRU (e.g., UE) may send the trained nested model / sub-model parameters (weights, gradients, number of training iterations, etc.), training dataset size, and corresponding trained sub-model ID to NW (S917).

[0281] The NW may aggregate the received nested model attributes into the global nested AI / ML model (S918).

[0282] The NW may send the updated nested AI / ML model to the WTRUs (e.g., UEs) (S919).

[0283] In FIG. 10, an example scenario is presented for the applications of the proposed nested AI / ML model architectures within the context of federated learning. The nested AI / ML model follows the form of Variation #1 shown in FIG. 5(a). Here, there are six WTRUs (e.g., UEs) and the NW may cluster them into three groups based on their static capabilities. For instance, one may consider the WTRU (e.g., UE) grouping based on the compute capability as follows.

[0284] WTRUi and WTRU2 have low compute capability and may be grouped in a first group. Hence, they are associated with Model A (i.e., the smallest sub-model).

[0285] WTRU3 and WTRU4 have medium compute capability and may be grouped in a second group. Then, they can run Model B.

[0286] WTRUs and WTRUe are configured with Model C (i.e., the full nested AI / ML model) since they have high compute capability and may be grouped in a third group. However, WTRUs and WTRUe prefer utilizing different versions of Model C with different quantization levels (please see Variation #5 shown in FIG. 6). Particularly, WTRUs has Model Ci, while WTRUe runs Model C2. On the other hand, they both update the model parameters for Model C.

[0287] The grouping criteria may not be limited to the compute capability (e.g., the number of Tx / Rx antennas and the number of sub-bands for the CSI prediction expressed in Exemplary Use Case #1).

[0288] As indicated in FIG. 9, the NW may transfer a particular sub-model with a particular submodel ID to the WTRUs (e.g., UEs). According to the federated learning, the WTRUs (e.g., UEs) may train all model parameters, or a part of the model based on dynamic conditions. For instance, UE5 and UE6 may train either Model A or Model B or Model C or Model B' or Model C. After the training, the WTRUs (e.g., UEs) may send the weight updates (or gradient updates) of the corresponding sub-model to the NW. Finally, the NW may aggregate all model parameters and feedback the updated nested models to the UEs.

[0289] Since all WTRUs (e.g., UEs) with different capabilities contribute the training of nested AI / ML model, they benefit from each other. For instance, even though UE3 has Model B, it can train (e.g., only) Model A since it is nested in Model B. Then, by sending the weight updates to the NW, UE3 may contribute to the improvement of Model A, which is preferred by WTRU1 and WTRU2.

[0290] The Nested AI / ML model architecture may enable the WTRUs (e.g., UEs) with diverse capabilities to collaboratively contribute to the training of an AI / ML model in the context of distributed / federated learning. Federated learning also has the advantage of alleviating the overhead of sending the local datasets to the NW while keeping the privacy and security of the data local to the UEs.

[0291] One or more examples described herein may be used separately or in combination for the WTRU to determine distributed / federated learning of nested models.

[0292] The WTRU (e.g., UE) may receive a request to provide its capabilities and / or dynamic condition to the NW. The WTRU (e.g., UE) may indicate, for example in UE-NR-Capability, its compute capabilities, functions(s) or component(s) for which an Al component can be configured with (e.g., online) learning capabilities, etc. The WTRU (e.g., UE) may indicate dynamic conditions e.g., WTRU (e.g., UE) power saving state, radio resources, QoS, latency, available compute resources, etc.

[0293] The WTRU (e.g., UE) may be configured to perform (e.g., online) learning for a given Al component with nested sub-model ID.

[0294] The WTRU (e.g., UE) may receive one or more of the following configurations: (1) one or more Al component identifier (e.g., Al component that may identify the function(s) being replaced by an AI / ML model); (2) one or more learning parameters (e.g., loss function, number oftraining iterations, loss threshold, learning method, etc.); or (3) one or more nested model attributes: (i) sub-model

[0295] The WTRU (e.g., UE) may receive an AI / ML nested model for the configured Al component.

[0296] The WTRU (e.g., UE) may receive an indication to start the (e.g., online) learning for the configured Al component and nested model sub-model ID.

[0297] The WTRU (e.g., UE) may apply (e.g., online) learning procedures for the configured Al component with nested model sub-model ID using local data.

[0298] The WTRU (e.g., UE) may transmit the trained AI / ML nested model parameters and number of training iterations corresponding to the nested model sub-model ID.

[0299] The WTRU (e.g., UE) may receive an updated AI / ML nested model for the configured Al component.

[0300] Methods and apparatus for hierarchical federated learning of nested models may be proposed.

[0301] The Nested AI / ML model architecture may enable the WTRUs (e.g., UEs) with diverse capabilities to collaboratively contribute to the training of an AI / ML model in the context of distributed / federated learning. Federated learning also has the advantage of alleviating the overhead of sending the local datasets to the NW while keeping the privacy and security of the data local to the WTRUs (e.g., UEs).Hierarchical federated learning (or decentralized federated learning) may enable grouping of WTRUs (e.g., UEs) with similar capabilities and / or dynamic conditions and uses Sidelink or WTRU-to-WTRU direct communication between WTRUs (e.g., UEs) to aggregate learned weights within Each (e.g., one or more) group or cluster of WTRUs (e.g., UEs).

[0302] A WTRU (e.g., UE) may be configured with grouping or clustering information where an aggregator WTRU (e.g., UE) identity is provided to Each (e.g., one or more) WTRU (e.g., UE) in the group or cluster of WTRUs (e.g., UEs). An aggregator WTRU (e.g., UE) may also be configured with the identity of other WTRUs (e.g., UEs) in the group. In a solution, WTRU (e.g., UE) identity of other WTRUs (e.g., UEs) in the group may be provided. Grouping information may change and the WTRU (e.g., UE) may receive updated grouping information or new aggregator WTRU (e.g., UE) identity. For example, this can be the case when a WTRU (e.g., UE) move out-of-coverage of the aggregator WTRU (e.g., UE). In this case, an alternate aggregator WTRU (e.g., UE) identity may be provided which is determined to be in coverage of the WTRU (e.g., UE).

[0303] In a solution, an aggregator WTRU (e.g., UE) may act as a relay augmenting the coverage and thus enabling the transmission of learned weights towards the NW and / or another WTRU (e.g., UE).

[0304] A WTRU (e.g., UE) may be configured with a nested model for an Al component. In a solution, the WTRU (e.g., UE) may be configured with a sub-model ID which may identify a part of the nested AI / ML model that is to be trained by the WTRU (e.g., UE) with using local data. In a solution, the configuration may include, but not limited to, learning parameters with number of training iterations, learning method, a loss function, nested model attributes with sub-model ID, performance thresholds.

[0305] Additionally, in a solution, the WTRU (e.g., UE) may be configured to select a sub-model ID based on performance thresholds. This configuration may enable the WTRU (e.g., UE) to perform training and inference with consideration for the WTRU (e.g., UE) static capabilities and / or dynamic conditions with using the same nested model.

[0306] In a solution, a WTRU (e.g., UE) with a configured nested model in the context of hierarchical federated learning, leverages the nested model architecture to enable the training of sub-models based on WTRU (e.g., UE) static capabilities and / or dynamic conditions. Moreover, a WTRU (e.g., UE) may be designated as an aggregator WTRU (e.g., UE) where the learned weights of the nested model from different WTRUs (e.g., UEs) can be aggregated as a whole or in part by a WTRU (e.g., UE).The aggregator WTRU (e.g., UE) along with other WTRUs (e.g., UEs) can form a group or a cluster of WTRUs (e.g., UEs) that may train on a common sub-model from the nested model architecture.

[0307] FIG. 11 illustrates a signal flow for the hierarchical federated training of nested AI / ML model architecture. Specifically, at Each (e.g., one or more) iteration of the hierarchical federated learning, the operation of nested models is depicted in FIG. 11, and it is summarized as follows.

[0308] The NW may request the capabilities of all WTRUs (e.g., UEs) (i.e., static capabilities and dynamic conditions) (step SI 110).

[0309] Each (e.g., one or more) WTRU (e.g., UE) may send its static capabilities and dynamic conditions to NW (step Si l l 1).

[0310] The NW may cluster the WTRUs (e.g., UEs) into groups based on their static capabilities and / or dynamic conditions. An aggregator WTRU (e.g., UE) is selected for Each (e.g., one or more) group or WTRUs (e.g., UEs). The NW may determine the nested AI / ML model architecture for a given Al component. The NW may map the WTRU (e.g., UE) groups to sub-models in the selected nested AI / ML model (step SI 112).

[0311] The NW may configure the aggregator WTRU (e.g., UE) with a list of WTRU (e.g., UE) identity in the group or cluster of WTRUs (e.g., UEs). The sub-model ID may be configured for the group (step SI 113).

[0312] The aggregator WTRU (e.g., UE) may transmit via Sidelink the aggregator WTRU (e.g., UE) identity to the WTRUs (e.g., UEs) in the group (step SI 114).

[0313] The NW may send the nested AI / ML model and the associated attributes, including but not limited to, sub-model ID, to the corresponding WTRUs (e.g., UEs) (step SI 115). The aggregator WTRU (e.g., UE) may receive the nested model with an indication to transmit the nested model to other WTRUs (e.g., UEs) in the group via Sidelink (step SI 116).

[0314] The NW may request the WTRU (e.g., UE) to start training of nested model sub-model ID with learning parameters like number of training iterations (step SI 117). The aggregator WTRU (e.g., UE) may indicate the start of training for a nested model sub-model ID and learning parameters to WTRUs (e.g., UEs) in the group (step SI 118).

[0315] The WTRU (e.g., UE) may acknowledge the training request (step SI 119).

[0316] Each (e.g., one or more) WTRU (e.g., UE) may train its associated nested model parameters using its own data set (step SI 120).

[0317] Each (e.g., one or more) WTRU (e.g., UE) may send the trained nested model parameters (weights, gradients, number of training iterations, etc.), training dataset size, trained sub-model ID to the aggregator WTRU (e.g., UE) (step SI 121).

[0318] The aggregator WTRU (e.g., UE) aggregate or combine the learned weights received from the WTRUs (e.g., UEs) in the group (step SI 122).

[0319] The aggregator WTRU (e.g., UE) may send the locally aggregated nested model attributes to the NW (step SI 123).

[0320] The NW may aggregate the received nested model attributes into the global nested AI / ML model (step SI 124).

[0321] The NW may send the updated nested AI / ML model to the WTRUs (e.g., UEs) (step S1 I25).

[0322] The aggregator WTRU (e.g., UE) may receive the updated nested model with an indication to transmit the nested model to other WTRUs (e.g., UEs) in the group via Sidelink (step SI 126).

[0323] FIG. 12 illustrates another federated learning example for the nested AI / ML model architecture. According to the hierarchical federated learning, there is a hierarchy among the WTRUs (e.g., UEs) for the model transfer and model training (i.e., weight update and model aggregation). The NW may cluster the WTRUs (e.g., UEs) into three groups based on their staticcapabilities and / or proximity. Furthermore, WTRU1 with Model A is chosen to assist kAWTRUs (e.g., UEs) with Model A (i.e., WTRU^ , WTRU1 / CA) via Sidelink or D2D communications. Also, WTRU2 with Model B assists kBWTRUs (e.g., UEs) with Model B and WTRU3 with Model C assists kcWTRUs (e.g., UEs) with Model C.

[0324] In addition to aggregating the full nested AI / ML model at the NW, the aggregation of Each (e.g., one or more) sub-model may be also done at WTRU1, WTRU2, and WTRU3 by means of the hierarchical federated learning. Hence, there may be two levels of model aggregation:

[0325] Sub-model aggregation: WTRU1, WTRU2, and WTRU3 may locally aggregate the corresponding sub-models. For instance, WTRU^i, WTRU1 2, ..., WTRUl fcmay individually train Model A, then, they may send the weight updates (or gradient updates) to UE1 via Sidelink or D2D communications. After receiving all of them, WTRU1 i.e., the aggregator WTRU (e.g., UE), may aggregate the updated weights (or gradients), then WTRU1 may transfer the updated Model A to the corresponding kAWTRUs (e.g., UEs) with Model A. A similar procedure may be applied for the aggregation of Model B at WTRU2 and Model C at WTRU3. The WTRU (e.g., UE) may use the number of training iterations information received from other WTRUs (e.g., UEs) in the group to perform a weighted model aggregation.

[0326] Also, WTRU2;I may send weight updates for Model A to WTRU 1 and / or weight updates for Model B' to WTRU2 via sidelink or D2D communications. Similarly, WTRU1 / CAmay send weight updates for Model A to WTRU2 and get Model A update from WTRU2.

[0327] Full model aggregation: WTRU1, WTRU2, and WTRU3 may send the updated Model A, Model B, and Model C, respectively, to the NW for the full nested AI / ML model aggregation. Then, the WTRUs (e.g., UEs) under different groups and using different sub-models collaboratively join the training of the full nested AI / ML model architecture.

[0328] Here, the frequency of sub-model aggregation and full model aggregation may be different. Therefore, it may result in reducing the AI / ML-related signaling overhead.

[0329] The Nested AI / ML model architecture may enable the WTRUs (e.g., UEs) with diverse capabilities to collaboratively contribute to the training of an AI / ML model in the context of distributed / federated learning. Federated learning also has the advantage of alleviating the overhead of sending the local datasets to the NW while keeping the privacy and security of the data local to the WTRUs (e.g., UEs).Hierarchical federated learning (or decentralized federated learning) may enable grouping of WTRUs (e.g., UEs) with similar capabilities and / or dynamic conditions and uses Sidelink or WTRU-to-WTRU direct communication between WTRUs (e.g., UEs) to aggregate learned weights within Each (e.g., one or more) group or cluster of WTRUs (e.g., UEs). A WTRU (e.g., UE) may be configured as an aggregator WTRU (e.g., UE) withinEach (e.g., one or more) group or cluster of WTRUs (e.g., UEs). The aggregator WTRU (e.g., UE) may transmit to the NW with the aggregated model parameters for a group or cluster of WTRUs (e.g., UEs).

[0330] One or more examples described herein may be used separately or in combination for the WTRU to determine hierarchical federated learning of nested models.

[0331] The WTRU (e.g., UE) may receive a request to provide its capabilities and / or dynamic condition to the NW. The WTRU (e.g., UE) may indicate, for example in UE-NR-Capability, its compute capabilities, functions(s) or component(s) for which an Al component can be configured with (e.g., online) learning capabilities, etc. The WTRU (e.g., UE) may indicate dynamic conditions e.g., WTRU (e.g., UE) power saving state, radio resources, QoS, latency, available compute resources, etc.

[0332] The WTRU (e.g., UE) may be configured to perform (e.g., online) learning for a given Al component with aggregator WTRU (e.g., UE) identity and / or identity of WTRUs (e.g., UEs) in the group or cluster of UEs. The Al component may identify the function(s) being replaced by the AI / ML nested model at the WTRU (e.g., UE). The aggregator WTRU (e.g., UE) identity may provide the WTRU (e.g., UE) with the WTRU (e.g., UE) identity to which AI / ML nested model parameters shall be sent to i.e., the WTRU (e.g., UE) configured for doing local aggregation via Sidelink or WTRU-to-WTRU direct communication within the configured group or cluster of WTRUs (e.g., UEs). If the WTRU (e.g., UE) is configured as the aggregator WTRU (e.g., UE), the WTRU (e.g., UE) may receive the identity of WTRUs (e.g., UEs) in the group or cluster of WTRUs (e g., UEs).

[0333] The WTRU (e.g., UE) may receive an AI / ML nested model for the configured Al component with (e.g., online) learning.

[0334] If the WTRU (e.g., UE) is configured as the aggregator WTRU (e.g., UE): (1) the WTRU (e.g., UE) may transmit an AI / ML nested model to the WTRUs (e.g., UEs) in the configured group or cluster of WTRUs (e.g., UEs) via Sidelink or WTRU-to-WTRU direct communication; and / or (2) the WTRU (e.g., UE) may transmit an indication to the WTRUs (e.g., UEs) in the configured group or cluster of WTRUs (e.g., UEs) via Sidelink or WTRU-to-WTRU direct communication to start the (e.g., online) learning of a configured Al component and nested model sub-model ID.

[0335] The WTRU (e.g., UE) may apply (e.g., online) learning procedures for the configured Al component using local data.

[0336] If the WTRU (e.g., UE) is configured as the aggregator WTRU (e.g., UE), the WTRU may perform any of the following actions: (1) the WTRU (e.g., UE) may receive via Sidelink or WTRU-to-WTRU direct communication, trained AI / ML nested model parameters withcorresponding sub-model ID from other WTRUs (e.g., UEs) in the configured group or cluster of WTRUs (e.g., UEs); (2) the WTRU (e.g., UE) may determine the nested model sub-model ID to perform aggregation on based on received trained nested model parameters and associated submodel IDs; or (3) the WTRU (e.g., UE) may transmit locally aggregated nested model parameters to the NW.

[0337] To demonstrate the utility of the disclosed method(s), we use the well-known MNIST dataset and train two sub-models to classify the MNIST digits. Referring to FIG. 5c but including (e.g., only) model A and B' and B, where the complexity (number of parameters) of A and B' is about the same. The complexity of the models is purposely made small (reflecting limited compute capability) and so that performance differences are easier to see. The WTRU supporting model A (e.g., only) has access to -10% of the MNIST dataset while WTRU supporting model B has access to the remaining -90% of the dataset. The models are fully connected with one hidden layer of size 10, 10 outputs. Model A 784 inputs. Model B 784+10 inputs. ReLu activations for all. Trained with Adam and learning rate = 0.01.

[0338] In training scenario 1 , model A is first trained, and the WTRU A may indicate that another WTRU may use model A outputs but does not wish to have its weights updated by other WTRUs. Model A is trained on its dataset using cross entropy loss function, and then the weights are frozen. Sub-model B' is then added to Model A to create model B. Recall that model B' is about the same capacity as model A and so would have the same performance if trained in the same way. WTRU B supporting model B (e.g., only) trains the model B' portion. In this case we wish to see if model B benefits for the outputs of model A. 20% of each dataset is reserved for testing. The inference results using the testing data are shown in first row of Table 1. Model A accuracy is 85.5% and Model B accuracy is 90.5%. Clearly, Model B benefits from this method even when model A does not fully collaborate.

[0339] In training scenario 2, both WTRUs may fully collaborate in training at the same time. As before, the WTRU supporting Model A has access to -10% of the MNIST dataset. Model A is updated by WTRU A as before, but now WTRU B also updates model A. WTRU B uses the sum of two cross entropy loss functions to update the parameters of model B, i.e., both sub-model A and sub-model B' are updated by WTRU B. Note that model A is expected to benefit from the increased access to data and model B expected to benefit from the extra capacity of model A. The inference results using the testing data are shown in row two of Table 1. Model A accuracy increased from 85.5% to 89.7% and Model B accuracy increased from 90.5% to 91.4% compared to training scenario 1. Model A in training scenario 1 is a baseline case since it does not make for the disclosed method(s) and can be used as point of reference for the other results.Table 1 Inference Results

[0340] According to certain embodiments, a WTRU may be configured with a Nested Al model applicable to one or more function(s) of a protocol stack e.g., as defined by [INPUT, OUTPUT, model architecture, bit-depth, sub-model IDs, etc.].

[0341] According to certain embodiments, a WTRU may receive signaling that may indicate the Nested Al model configuration with its attributes and / or sub-model IDs. The activation of a submodel ID may be a configuration aspect of the WTRU. A WTRU may be configured with one or more Nested Al model(s) (possibly acquired via unicast, broadcast or multicast transmissions).

[0342] According to certain embodiments, a WTRU may determine a sub-model from the Nested Al model based on dynamic conditions (e.g., UE power saving state, radio resources, QoS, latency, available compute resources, available data observation, etc.).

[0343] According to certain embodiments, a WTRU may activate a selected sub-model of the Nested Al model associated with an Al component.

[0344] According to certain embodiments, a WTRU may transmit an indication of selected submodel for a given Nested Al model.

[0345] According to certain embodiments, a WTRU may be configured with one or more Nested Al model(s) (possibly acquired via unicast, broadcast or multicast transmissions).

[0346] According to certain embodiments, a WTRU may activate a sub-model of a Nested Al model based on received configuration.

[0347] According to certain embodiments, a WTRU may transmit performance metrics or KPIs associated to dynamic conditions (e.g., UE power saving state, radio resources, QoS, latency, available compute resources, available data observation, etc.).

[0348] According to certain embodiments, a WTRU may receive a configuration update for a Nested Al model.

[0349] According to certain embodiments, a WTRU may activate a sub-model of a Nested Al model based on the configuration update.

[0350] According to certain embodiments, a WTRU may transmit an acknowledgement of selected sub-model for a given Nested Al model.

[0351] According to certain embodiments, a group or cluster of WTRUs may be configured with one or more Nested Al model(s) (possibly acquired via multicast, unicast, or broadcast transmissions), where Each (e.g., one or more) WTRU may train the nested model(s) based on itsdevice capabilities and / or dynamic conditions and update the NW with trained nested model(s) parameters. The NW may aggregate the trained nested models and then may configure the WTRUs with the updated nested models (possibly via multicast, unicast, or broadcast transmissions).

[0352] According to certain embodiments, a WTRU may perform local aggregation of learned nested model parameters received from other WTRU via sidelink. The aggregator WTRU may configure and coordinate the training of nested model within the group or cluster of WTRUs. The aggregator WTRU may transmit aggregated model parameters to the NW for global nested model update.

[0353] FIG. 13 is a flowchart illustrating a representative method 1300 implemented by a WTRU 102. Referring to FIG. 13, the representative method 1300 may include, at block 1310, sending, to a network node, first information indicating one or more (e.g., static) capabilities and / or one or more (e.g., dynamic) conditions associated to the WTRU. At block 1320, the representative method 1300 may include receiving configuration information for training a nested AI / ML model based on the first information, wherein the nested AI / ML model comprises a plurality of AI / ML sub-models. At block 1330, the representative method 1300 may include performing training for at least an AI / ML sub-model based on the configuration information. At block 1340, the representative method 1300 may include sending second information indicating the trained AI / ML sub-model parameters and number of training iterations.

[0354] According to certain embodiments, the configuration information may indicate any of (1) an identifier associated to the AI / ML model, (2) learning parameters, (3) an identifier associated to the at least AI / ML sub-model, and (4) at least one attribute associated to the at least AI / ML sub-model.

[0355] According to certain embodiments, the configuration information may indicate an identity associated to a WTRU configured for local aggregation of trained AI / ML sub-model parameters, and may further include sending the second information to the WTRU configured for local aggregation.

[0356] According to certain embodiments, the configuration information may indicate any of identities associated to a plurality of WTRUs.

[0357] According to certain embodiments, the representative method 1300 may further comprise any of receiving from the plurality of WTRUs, trained AI / ML sub-model parameters; aggregating locally the trained AI / ML sub-model parameters received; and sending, to the network node, the locally aggregated trained AI / ML sub-model parameters received.

[0358] According to certain embodiments, the representative method 1300 may further comprise: sending the AI / ML nested model to the plurality of WTRUs via sidelink.

[0359] According to certain embodiments, the representative method 1300 may further comprise from the plurality of WTRUs, trained AI / ML sub-model parameters via sidelink.

[0360] According to certain embodiments, the capabilities associated to the WTRU may comprise any of: number of antennas, computing capability, storage capability.

[0361] According to certain embodiments, the dynamic conditions associated to the WTRU may comprise any of: power saving state, radio resources, QoS, latency, available compute resources, link conditions, channel types.

[0362] FIG. 14 is a flowchart illustrating a representative method 1400 implemented by a first WTRU 102.

[0363] Referring to FIG. 14, the representative method 1400 may include, at block 1410, sending, to a network node, first information indicating one or more (e.g., static) capabilities and / or one or more (e.g., dynamic) conditions associated with the first WTRU.

[0364] At block 1420, the representative method 1400 may include receiving, from the network node, based on the first information, configuration information for training a nested AI / ML model, wherein the nested AI / ML model comprises a plurality of AI / ML sub-models.

[0365] At block 1430, the representative method 1400 may include performing, based on the configuration information, a training of at least one AI / ML sub-model from the plurality of AI / ML sub-models.

[0366] At block 1440, the representative method 1400 may include sending, to the network node, second information indicating any of: (1) a first set of model parameters associated with the training of the at least one AI / ML sub-model, and (2) a number of training iterations associated with the training of the at least one AI / ML sub-model.

[0367] According to certain embodiments, the configuration information may indicate any of: (1) an identifier associated with the nested AI / ML model, (2) one or more learning parameters, (3) one or more identifiers associated with the at least one AI / ML sub-model, and (4) at least one attribute associated with the at least one AI / ML sub-model.

[0368] According to certain embodiments, the configuration information may indicate an identifier associated with a second WTRU configured for local aggregation of AI / ML sub-model parameters, and further comprising: sending the second information to the second WTRU.

[0369] According to certain embodiments, the configuration information may indicate an identifier associated with a third WTRU, and According to certain embodiments, the representative method 1400 may further comprise any of: receiving, from the third WTRU, a second set of model parameters associated with the training by the third WTRU of the at least one AI / ML submodel; obtaining a third set of model parameters by aggregating locally the first set of modelparameters and the second set of model parameters and sending, to the network node, the third set of model parameters.

[0370] According to certain embodiments, the representative method 1400 may further comprise sending, to the third WTRU via a sidelink path, the third set of model parameters.

[0371] According to certain embodiments, the second set of model parameters are received from the third WTRU via the sidelink path.

[0372] According to certain embodiments, the (e.g., static) capabilities associated with the first WTRU may indicate any of: one or more number of antennas associated with the first WTRU, one or more computing capability associated with the first WTRU, one or more storage capability associated with the first WTRU.

[0373] According to certain embodiments, the (e.g., dynamic) conditions associated with the first WTRU may indicate any of: (1) a power saving state associated with the first WTRU, (2) one or more radio resources associated with a communication link of the first WTRU, (3) one or more quality of service associated with the communication link of the first WTRU, (4) a latency value associated with the communication link of the first WTRU, (5) one or more available compute resources associated with the first WTRU, (6) one or more link conditions associated with the communication link of the first WTRU, and (7) one or more channel types associated with the communication link of the first WTRU.

[0374] According to certain embodiments, the at least one AI / ML sub-model is selected based on the (e.g., dynamic) conditions.

[0375] FIG. 15 is a flowchart illustrating a representative method 1500 implemented by a first WTRU 102.

[0376] Referring to FIG. 15, the representative method 1500 may include, at block 1510, receiving, from a plurality of WTRUs, first information indicating one or more capabilities and / or one or more conditions associated with each WTRU of the plurality of WTRUs.

[0377] At block 1520, the representative method 1500 may include sending, to the plurality of WTRUs, based on the first information, configuration information for training a nested AI / ML model, wherein the nested AI / ML model comprises a plurality of AI / ML sub-models.

[0378] At block 1530, the representative method 1500 may include receiving, from the plurality of WTRUs, a plurality of second information, indicating any of: (1) a first set of model parameters associated with the training of at least one AI / ML sub-model from the plurality of AI / ML submodels by each WTRU of the plurality of WTRUs, and (2) a number of training iterations associated with the training of the at least one AI / ML sub-model by each WTRU of the plurality of WTRUs.

[0379] According to certain embodiments, the representative method 1500 may further comprise obtaining a second set of model parameters associated with the nested AI / ML model by aggregating the first set of model parameters from the plurality of second information; and sending, to the plurality of WTRUs, the second set of model parameters associated with the nested AI / ML model.

[0380] According to certain embodiments, the configuration information may indicate any of: (1) an identifier associated with the nested AI / ML model, (2) one or more learning parameters, (3) one or more identifiers associated with the at least one AI / ML sub-model, and (4) at least one attribute associated with the at least one AI / ML sub-model.

[0381] According to certain embodiments, the plurality of second information is received from the plurality of WTRUs via a plurality of sidelink paths.

[0382] According to certain embodiments, the second set of model parameters associated with the nested AI / ML model is sent to the plurality of WTRUs via the plurality of sidelink paths.

[0383] According to certain embodiments, the configuration information may indicate an identifier associated with a first WTRU of the plurality of WTRUs, wherein the first WTRU is configured for local aggregation of AI / ML sub-model parameters, and further comprising: receiving, from the first WTRU, third set of model parameters, wherein the third set of model parameters are aggregated from a plurality of set of model parameters obtained from the plurality of WTRUs.

[0384] According to certain embodiments, the capabilities associated with the first WTRU may indicate any of: one or more number of antennas associated with the first WTRU, one or more computing capability associated with the first WTRU, one or more storage capability associated with the first WTRU.

[0385] According to certain embodiments, the conditions associated with the first WTRU may indicate any of: (1) a power saving state associated with the first WTRU, (2) one or more radio resources associated with a communication link of the first WTRU, (3) one or more quality of service associated with the communication link of the first WTRU, (4) a latency value associated with the communication link of the first WTRU, (5) one or more available compute resources associated with the first WTRU, (6) one or more link conditions associated with the communication link of the first WTRU, and (7) one or more channel types associated with the communication link of the first WTRU.

[0386] Although features and elements are provided above in particular combinations, one of ordinary skill in the art will appreciate that each feature or element can be used alone or in any combination with the other features and elements. The present disclosure is not to be limited interms of the particular embodiments described in this application, which are intended as illustrations of various aspects. Many modifications and variations may be made without departing from its spirit and scope, as will be apparent to those skilled in the art. No element, act, or instruction used in the description of the present application should be construed as critical or essential to the invention unless explicitly provided as such. Functionally equivalent methods and apparatuses within the scope of the disclosure, in addition to those enumerated herein, will be apparent to those skilled in the art from the foregoing descriptions. Such modifications and variations are intended to fall within the scope of the appended claims. The present disclosure is to be limited only by the terms of the appended claims, along with the full scope of equivalents to which such claims are entitled. It is to be understood that this disclosure is not limited to particular methods or systems.

[0387] The foregoing embodiments are discussed, for simplicity, with regard to the terminology and structure of infrared capable devices, i.e., infrared emitters and receivers. However, the embodiments discussed are not limited to these systems but may be applied to other systems that use other forms of electromagnetic waves or non-electromagnetic waves such as acoustic waves.

[0388] It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting. As used herein, the term "video" or the term "imagery" may mean any of a snapshot, single image and / or multiple images displayed over a time basis. As another example, when referred to herein, the terms "user equipment" and its abbreviation "UE", the term "remote" and / or the terms "head mounted display" or its abbreviation "HMD" may mean or include (i) a wireless transmit and / or receive unit (WTRU); (ii) any of a number of embodiments of a WTRU; (iii) a wireless-capable and / or wired-capable (e.g., tetherable) device configured with, inter alia, some or all structures and functionality of a WTRU; (iii) a wireless-capable and / or wired-capable device configured with less than all structures and functionality of a WTRU; or (iv) the like. Details of an example WTRU, which may be representative of any WTRU recited herein, are provided herein with respect to FIGs. 1 A-1D. As another example, various disclosed embodiments herein supra and infra are described as utilizing a head mounted display. Those skilled in the art will recognize that a device other than the head mounted display may be utilized and some or all of the disclosure and various disclosed embodiments can be modified accordingly without undue experimentation. Examples of such other device may include a drone or other device configured to stream information for providing the adapted reality experience.

[0389] In addition, the methods provided herein may be implemented in a computer program, software, or firmware incorporated in a computer-readable medium for execution by a computeror processor. Examples of computer-readable media include electronic signals (transmitted over wired or wireless connections) and computer-readable storage media. Examples of computer- readable storage media include, but are not limited to, a read only memory (ROM), a random access memory (RAM), a register, cache memory, semiconductor memory devices, magnetic media such as internal hard disks and removable disks, magneto-optical media, and optical media such as CD-ROM disks, and digital versatile disks (DVDs). A processor in association with software may be used to implement a radio frequency transceiver for use in a WTRU, UE, terminal, base station, RNC, or any host computer.

[0390] Variations of the method, apparatus and system provided above are possible without departing from the scope of the invention. In view of the wide variety of embodiments that can be applied, it should be understood that the illustrated embodiments are examples only, and should not be taken as limiting the scope of the following claims. For instance, the embodiments provided herein include handheld devices, which may include or be utilized with any appropriate voltage source, such as a battery and the like, providing any appropriate voltage.

[0391] Moreover, in the embodiments provided above, processing platforms, computing systems, controllers, and other devices that include processors are noted. These devices may include at least one Central Processing Unit ("CPU") and memory. In accordance with the practices of persons skilled in the art of computer programming, reference to acts and symbolic representations of operations or instructions may be performed by the various CPUs and memories. Such acts and operations or instructions may be referred to as being "executed," "computer executed" or "CPU executed."

[0392] One of ordinary skill in the art will appreciate that the acts and symbolically represented operations or instructions include the manipulation of electrical signals by the CPU. An electrical system represents data bits that can cause a resulting transformation or reduction of the electrical signals and the maintenance of data bits at memory locations in a memory system to thereby reconfigure or otherwise alter the CPU's operation, as well as other processing of signals. The memory locations where data bits are maintained are physical locations that have particular electrical, magnetic, optical, or organic properties corresponding to or representative of the data bits. It should be understood that the embodiments are not limited to the above-mentioned platforms or CPUs and that other platforms and CPUs may support the provided methods.

[0393] The data bits may also be maintained on a computer readable medium including magnetic disks, optical disks, and any other volatile (e.g., Random Access Memory (RAM)) or non-volatile (e.g., Read-Only Memory (ROM)) mass storage system readable by the CPU. The computer readable medium may include cooperating or interconnected computer readable medium, whichexist exclusively on the processing system or are distributed among multiple interconnected processing systems that may be local or remote to the processing system. It should be understood that the embodiments are not limited to the above-mentioned memories and that other platforms and memories may support the provided methods.

[0394] In an illustrative embodiment, any of the operations, processes, etc. described herein may be implemented as computer-readable instructions stored on a computer-readable medium. The computer-readable instructions may be executed by a processor of a mobile unit, a network element, and / or any other computing device.

[0395] There is little distinction left between hardware and software implementations of aspects of systems. The use of hardware or software is generally (but not always, in that in certain contexts the choice between hardware and software may become significant) a design choice representing cost versus efficiency tradeoffs. There may be various vehicles by which processes and / or systems and / or other technologies described herein may be effected (e.g., hardware, software, and / or firmware), and the preferred vehicle may vary with the context in which the processes and / or systems and / or other technologies are deployed. For example, if an implementer determines that speed and accuracy are paramount, the implementer may opt for a mainly hardware and / or firmware vehicle. If flexibility is paramount, the implementer may opt for a mainly software implementation. Alternatively, the implementer may opt for some combination of hardware, software, and / or firmware.

[0396] The foregoing detailed description has set forth various embodiments of the devices and / or processes via the use of block diagrams, flowcharts, and / or examples. Insofar as such block diagrams, flowcharts, and / or examples include one or more functions and / or operations, it will be understood by those within the art that each function and / or operation within such block diagrams, flowcharts, or examples may be implemented, individually and / or collectively, by a wide range of hardware, software, firmware, or virtually any combination thereof. In an embodiment, several portions of the subject matter described herein may be implemented via Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), digital signal processors (DSPs), and / or other integrated formats. However, those skilled in the art will recognize that some aspects of the embodiments disclosed herein, in whole or in part, may be equivalently implemented in integrated circuits, as one or more computer programs running on one or more computers (e.g., as one or more programs running on one or more computer systems), as one or more programs running on one or more processors (e.g., as one or more programs running on one or more microprocessors), as firmware, or as virtually any combination thereof, and that designing the circuitry and / or writing the code for the software and or firmware would be well within the skillof one of skill in the art in light of this disclosure. In addition, those skilled in the art will appreciate that the mechanisms of the subject matter described herein may be distributed as a program product in a variety of forms, and that an illustrative embodiment of the subject matter described herein applies regardless of the particular type of signal bearing medium used to actually carry out the distribution. Examples of a signal bearing medium include, but are not limited to, the following: a recordable type medium such as a floppy disk, a hard disk drive, a CD, a DVD, a digital tape, a computer memory, etc., and a transmission type medium such as a digital and / or an analog communication medium (e.g., a fiber optic cable, a waveguide, a wired communications link, a wireless communication link, etc.).

[0397] Those skilled in the art will recognize that it is common within the art to describe devices and / or processes in the fashion set forth herein, and thereafter use engineering practices to integrate such described devices and / or processes into data processing systems. That is, at least a portion of the devices and / or processes described herein may be integrated into a data processing system via a reasonable amount of experimentation. Those having skill in the art will recognize that a typical data processing system may generally include one or more of a system unit housing, a video display device, a memory such as volatile and non-volatile memory, processors such as microprocessors and digital signal processors, computational entities such as operating systems, drivers, graphical user interfaces, and applications programs, one or more interaction devices, such as a touch pad or screen, and / or control systems including feedback loops and control motors (e.g., feedback for sensing position and / or velocity, control motors for moving and / or adjusting components and / or quantities). A typical data processing system may be implemented utilizing any suitable commercially available components, such as those typically found in data computing / communication and / or network computing / communication systems.

[0398] The herein described subject matter sometimes illustrates different components included within, or connected with, different other components. It is to be understood that such depicted architectures are merely examples, and that in fact many other architectures may be implemented which achieve the same functionality. In a conceptual sense, any arrangement of components to achieve the same functionality is effectively "associated" such that the desired functionality may be achieved. Hence, any two components herein combined to achieve a particular functionality may be seen as "associated with" each other such that the desired functionality is achieved, irrespective of architectures or intermedial components. Likewise, any two components so associated may also be viewed as being "operably connected", or "operably coupled", to each other to achieve the desired functionality, and any two components capable of being so associated may also be viewed as being "operably couplable" to each other to achieve the desired functionality.Specific examples of operably couplable include but are not limited to physically mateable and / or physically interacting components and / or wirelessly interactable and / or wirelessly interacting components and / or logically interacting and / or logically interactable components.

[0399] With respect to the use of substantially any plural and / or singular terms herein, those having skill in the art can translate from the plural to the singular and / or from the singular to the plural as is appropriate to the context and / or application. The various singular / plural permutations may be expressly set forth herein for sake of clarity.

[0400] It will be understood by those within the art that, in general, terms used herein, and especially in the appended claims (e.g., bodies of the appended claims) are generally intended as "open" terms (e.g., the term "including" should be interpreted as "including but not limited to," the term "having" should be interpreted as "having at least," the term "includes" should be interpreted as "includes but is not limited to," etc.). It will be further understood by those within the art that if a specific number of an introduced claim recitation is intended, such an intent will be explicitly recited in the claim, and in the absence of such recitation no such intent is present. For example, where only one item is intended, the term "single" or similar language may be used. As an aid to understanding, the following appended claims and / or the descriptions herein may include usage of the introductory phrases "at least one" and "one or more" to introduce claim recitations. However, the use of such phrases should not be construed to imply that the introduction of a claim recitation by the indefinite articles "a" or "an" limits any particular claim including such introduced claim recitation to embodiments including only one such recitation, even when the same claim includes the introductory phrases "one or more" or "at least one" and indefinite articles such as "a" or "an" (e.g., "a" and / or "an" should be interpreted to mean "at least one" or "one or more"). The same holds true for the use of definite articles used to introduce claim recitations. In addition, even if a specific number of an introduced claim recitation is explicitly recited, those skilled in the art will recognize that such recitation should be interpreted to mean at least the recited number (e.g., the bare recitation of "two recitations," without other modifiers, means at least two recitations, or two or more recitations). Furthermore, in those instances where a convention analogous to "at least one of A, B, and C, etc." is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., "a system having at least one of A, B, and C" would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together, etc.). In those instances where a convention analogous to "at least one of A, B, or C, etc." is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., "a system having at least one of A, B, or C" would include but not be limited to systems thathave A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together, etc.). It will be further understood by those within the art that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the description, claims, or drawings, should be understood to contemplate the possibilities of including one of the terms, either of the terms, or both terms. For example, the phrase "A or B" will be understood to include the possibilities of "A" or "B" or "A and B." Further, the terms "any of' followed by a listing of a plurality of items and / or a plurality of categories of items, as used herein, are intended to include "any of," "any combination of," "any multiple of," and / or "any combination of multiples of the items and / or the categories of items, individually or in conjunction with other items and / or other categories of items. Moreover, as used herein, the term "set" is intended to include any number of items, including zero. Additionally, as used herein, the term "number" is intended to include any number, including zero. And the term "multiple", as used herein, is intended to be synonymous with "a plurality".

[0401] In addition, where features or aspects of the disclosure are described in terms of Markush groups, those skilled in the art will recognize that the disclosure is also thereby described in terms of any individual member or subgroup of members of the Markush group.

[0402] As will be understood by one skilled in the art, for any and all purposes, such as in terms of providing a written description, all ranges disclosed herein also encompass any and all possible subranges and combinations of subranges thereof. Any listed range can be easily recognized as sufficiently describing and enabling the same range being broken down into at least equal halves, thirds, quarters, fifths, tenths, etc. As a non-limiting example, each range discussed herein may be readily broken down into a lower third, middle third and upper third, etc. As will also be understood by one skilled in the art all language such as "up to," "at least," "greater than," "less than," and the like includes the number recited and refers to ranges which can be subsequently broken down into subranges as discussed above. Finally, as will be understood by one skilled in the art, a range includes each individual member. Thus, for example, a group having 1-3 cells refers to groups having 1, 2, or 3 cells. Similarly, a group having 1-5 cells refers to groups having 1, 2, 3, 4, or 5 cells, and so forth.

[0403] Moreover, the claims should not be read as limited to the provided order or elements unless stated to that effect. In addition, use of the terms "means for" in any claim is intended to invoke 35 U.S.C. §112, 6 or means-plus-function claim format, and any claim without the terms "means for" is not so intended.

Claims

CLAIMSWhat is claimed is:

1. A method implemented by a first wireless transmit / receive unit (WTRU), the method comprising: sending, to a network node, first information indicating at least one of one or more capabilities and one or more conditions associated with the first WTRU; receiving, from the network node, based on the first information, configuration information for training a nested artificial intelligence / machine learning (AI / ML) model, wherein the nested AI / ML model comprises a plurality of AI / ML sub-models; performing, based on the configuration information, a training of at least one AI / ML submodel from the plurality of AI / ML sub-models; and sending, to the network node, second information indicating any of: (1) a first set of model parameters associated with the training of the at least one AI / ML sub-model, and (2) a number of training iterations associated with the training of the at least one AI / ML sub-model.

2. The method according to claim 1, wherein the configuration information indicates any of: (1) an identifier associated with the nested AI / ML model, (2) one or more learning parameters, (3) one or more identifiers associated with the at least one AI / ML sub-model, and (4) at least one attribute associated with the at least one AI / ML sub-model.

3. The method according to any of claims 1-2, wherein the configuration information indicates an identifier associated with a second WTRU configured for local aggregation of AI / ML sub-model parameters, and further comprising: sending a third information to the second WTRU indicating any of: (1) a first set of model parameters associated with the training of the at least one AI / ML sub-model, and (2) a number of training iterations associated with the training of the at least one AI / ML sub-model.

4. The method according to any of claims 1-2, wherein the configuration information indicates an identifier associated with a third WTRU, and wherein the method further comprises: receiving, from the third WTRU, a second set of model parameters associated with the training by the third WTRU of the at least one AI / ML sub-model; obtaining a third set of model parameters by aggregating locally the first set of model parameters and the second set of model parameters; and sending, to the network node, the third set of model parameters.

5. The method according to claim 4, further comprising: sending, to the third WTRU via a sidelink path, the third set of model parameters.

6. The method according to claim 5, wherein the second set of model parameters are received from the third WTRU via the sidelink path.

7. The method according to any of claims 1-6, wherein the one or more capabilities associated with the first WTRU indicate any of: one or more number of antennas associated with the first WTRU, one or more computing capability associated with the first WTRU, and one or more storage capability associated with the first WTRU.

8. The method according to any of claims 1-7, wherein the one or more conditions associated with the first WTRU indicate any of: (1) a power saving state associated with the first WTRU, (2) one or more radio resources associated with a communication link of the first WTRU, (3) one or more quality of service associated with the communication link of the first WTRU, (4) a latency value associated with the communication link of the first WTRU, (5) one or more available compute resources associated with the first WTRU, (6) one or more link conditions associated with the communication link of the first WTRU, and (7) one or more channel types associated with the communication link of the first WTRU.

9. The method according to any of claims 1-8, wherein the at least one AI / ML sub-model is selected based on the conditions.

10. A method implemented by a network node, the method comprising: receiving, from a plurality of wireless transmit / receive units (WTRUs), first information indicating at least one of one or more capabilities and one or more conditions associated with each WTRU of the plurality of WTRUs; sending, to the plurality of WTRUs, based on the first information, configuration information for training a nested artificial intelligence / machine learning (AI / ML) model, wherein the nested AI / ML model comprises a plurality of AI / ML sub-models; and receiving, from the plurality of WTRUs, a plurality of second information, indicating any of: (1) a first set of model parameters associated with the training of at least one AI / ML sub-model from the plurality of AI / ML sub-models by each WTRU of the plurality of WTRUs, and (2) a number of training iterations associated with the training of the at least one AI / ML sub-model by each WTRU of the plurality of WTRUs.

11. The method according to claim 10, further comprising:obtaining a second set of model parameters associated with the nested AI / ML model by aggregating the first set of model parameters from the plurality of second information; and sending, to the plurality of WTRUs, the second set of model parameters associated with the nested AI / ML model.

12. The method according to any of claims 10-11, wherein the configuration information indicates any of: (1) an identifier associated with the nested AI / ML model, (2) one or more learning parameters, (3) one or more identifiers associated with the at least one AI / ML sub-model, and (4) at least one attribute associated with the at least one AI / ML sub-model.

13. The method according to any of claims 10-12, wherein the plurality of second information is received from the plurality of WTRUs via a plurality of sidelink paths.

14. The method according to claim 13, wherein the second set of model parameters associated with the nested AI / ML model is sent to the plurality of WTRUs via the plurality of sidelink paths.

15. The method according to any of claims 10-14, wherein the configuration information indicates an identifier associated with a first WTRU of the plurality of WTRUs, wherein the first WTRU is configured for local aggregation of AI / ML sub-model parameters, and further comprising: receiving, from the first WTRU, third set of model parameters, wherein the third set of model parameters are aggregated from a plurality of set of model parameters obtained from the plurality of WTRUs.

16. The method according to any of claims 11-15, wherein the one or more capabilities associated with the first WTRU indicate any of: one or more number of antennas associated with the first WTRU, one or more computing capability associated with the first WTRU, and one or more storage capability associated with the first WTRU.

17. The method according to any of claims 10-16, wherein the one or more conditions associated with the first WTRU indicate any of: (1) a power saving state associated with the first WTRU, (2) one or more radio resources associated with a communication link of the first WTRU, (3) one or more quality of service associated with the communication link of the first WTRU, (4) a latency value associated with the communication link of the first WTRU, (5) one or more available compute resources associated with the first WTRU, (6) one or more link conditions associated with the communication link of the first WTRU, and (7) one or more channel types associated with the communication link of the first WTRU.

18. A wireless transmit / receive unit (WTRU) comprising circuitry, including a transmitter, a receiver, a processor and memory, the WTRU configured to: send, to a network node, first information indicating at least one of one or more capabilities and one or more conditions associated with the WTRU; receive, from the network node, based on the first information, configuration information for training a nested artificial intelligence / machine learning (AI / ML) model, wherein the nested AI / ML model comprises a plurality of AI / ML sub-models; perform, based on the configuration information, a training of at least one AI / ML sub-model from the plurality of AI / ML sub-models; and send, to the network node, second information indicating any of (1) a first set of model parameters associated with the training of the at least one AI / ML sub-model, and (2) a number of training iterations associated with the training of the at least one AI / ML sub-model.

19. The WTRU according to claim 18, wherein the configuration information indicates any of (1) an identifier associated with the nested AI / ML model, (2) one or more learning parameters, (3) one or more identifiers associated with the at least one AI / ML sub-model, and (4) at least one attribute associated with the at least one AI / ML sub-model.

20. A network node, comprising circuitry, including a transmitter, a receiver, a processor and memory, the network node configured to: receive, from a plurality of wireless transmit / receive units (WTRUs), first information indicating at least one of one or more capabilities and one or more conditions associated with each WTRU of the plurality of WTRUs; send, to the plurality of WTRUs, based on the first information, configuration information for training a nested artificial intelligence / machine learning (AI / ML) model, wherein the nested AI / ML model comprises a plurality of AI / ML sub-models; and receive, from the plurality of WTRUs, a plurality of second information, indicating any of: (1) a first set of model parameters associated with the training of at least one AI / ML sub-model from the plurality of AI / ML sub-models by each WTRU of the plurality of WTRUs, and (2) a number of training iterations associated with the training of the at least one AI / ML sub-model by each WTRU of the plurality of WTRUs.

Images