Wireless communication system and method for artificial intelligence / machine learning data transmission

Abstract

A method for realizing AI / ML data transmission between a first node and a second node based on UP in a wireless communication system comprises the following steps: receiving AI data information and / or converting the AI data information into an AI data stream through an AI data adaptation protocol ADAP sub-layer, the AI data information being forced to be executed as the AI data stream according to the characteristics of the AI data information, and the ADAP sublayer maps the AI data stream into a quality of service (QoS) stream and / or a radio bearer (RB) (a control bearer (SRB) and / or a data bearer (DRB)), and / or a radio link control (RLC) channel / bearer, and / or a logic channel.

Description

Technical Field

[0001] This disclosure relates to the field of wireless communication systems, and more particularly to a wireless communication system and method for delivering / transmitting artificial intelligence (AI) / machine learning (ML) data between a user equipment (UE) and a network (e.g., gNB, LMF, etc.) via a user plane (UP). More specifically, the objective is to transmit AI data using a UP-based method. Compared to control plane (CP)-based methods, this disclosure reduces signaling overhead and transmission conflicts in wireless communication. Background Technology

[0002] For AI / ML research in aerial inference, the general framework agreed upon in RAN3 TR 37.817 is considered a starting point, such as... Figure 1 As shown, this indicates that data collection is beneficial for certain AI operations, such as model training and model inference. Following the progress of the conference and a comprehensive discussion of over-the-air inference, the data collection findings were concluded at the RAN1#110bis meeting as follows:

[0003] in conclusion: In lifecycle management (LCM), data collection can be used for various purposes, such as model training, model inference, model monitoring, model selection, and model updates. Each purpose may have different requirements and potential normative impacts.

[0004] In the follow-up email discussion to RAN2#120, many companies agreed to research data collection solutions for AI / ML models on both the network NW and UE sides. Furthermore, domain experts summarized several existing frameworks and requested an analysis of whether these legacy frameworks could be reused or enhanced for data collection. The existing frameworks are listed below:

[0005] -Mobile Data Collection (MDT); -UE assistance information (defined in the RRC specification); -Early idle / inactive state measurement; -RRC measurement report; - Channel State Information (CSI) reporting framework; and -LPP provides location information.

[0006] At the RAN2#121 meeting, the preliminary analysis of the data collection framework provided by Apple Inc. (Tdoc[23002286]) was identified as the starting point, the contents of which are shown in Table 1.

[0007] Table 1: Preliminary analysis of the existing data collection framework.

[0008] As mentioned above, data transmission / transmission also involves two directions: one is the NW transmitting data to the UE, and the other is the UE transmitting data to the NW, as shown in Figure 2. It is evident that the existing framework described above is suitable for uplink data collection. Furthermore, all these existing frameworks are based on the CP method, whose maximum data size is limited to 9KB or 144KB for UL messages (if segmentation is allowed), and 45KB for DL ​​messages. However, the dataset size can be large; for example, the data size used for model training can reach at least tens of MB, or even GB. Meanwhile, this issue was also addressed in contributions from several companies at the latest RAN1#112 meeting, with details as follows:

[0009] [R1-2301586 MediaTek]: "For CSI compressed training type 3 (training separately on the NW side and the UE side), we should consider how to exchange the dataset... If the dataset exchange is carried out in accordance with the 3GPP specifications, given the large size of the dataset, we need to pay attention to the air interface overhead caused by the dataset exchange."

[0010] Therefore, using a CP-based method for data transmission will lead to increased overhead and significantly prolonged latency. Therefore, a UP-based method should be investigated for data transmission / transmission.

[0011] In existing patents or academic papers, there are relatively few prior technologies concerning AI / ML data delivery / transfer in wireless communication systems. Therefore, based on the above discussion, it is necessary to conduct in-depth research on the methods of transmitting / transferring artificial intelligence (AI) / machine learning (ML) data in different application scenarios in the future. Summary of the Invention

[0012] The purpose of this disclosure is to propose a wireless communication system and a method for transmitting AI / ML data between a UE and a NW (such as a gNB, LMF, etc.). This method aims to transmit AI data using a UP-based approach, which is not yet addressed in current research and reduces signaling overhead and transmission conflicts compared to CP-based methods.

[0013] In a first aspect of this disclosure, a method for transmitting AI / ML data between a first node and a second node based on UP in a wireless communication system is provided, comprising: receiving AI data information and / or converting the AI ​​data information into an AI data stream through an AI Data Adaptation Protocol (ADAP) sublayer, wherein the AI ​​data information is forcibly converted into the AI ​​data stream according to its characteristics, and the ADAP sublayer maps the AI ​​data stream to a Quality of Service (QoS) stream and / or a Radio Bearer (RB) (Control Bearer (SRB) and / or Data Bearer (DRB)), and / or a Radio Link Control (RLC) channel / bearer, and / or a logical channel.

[0014] In a second aspect of this disclosure, a method is provided for transmitting artificial intelligence / machine learning (AI / ML) data between a first node and a second node based on a UP in a wireless communication system. The method comprises: indicating / distinguishing AI data information / AI data streams through an AI data stream identifier (ADFI); and / or indicating a QoS stream carrying the AI ​​data information through a Quality of Service (QFI), so that the AI ​​data stream is mapped to a conventional QoS stream.

[0015] In a third aspect of this disclosure, a method is provided for transmitting artificial intelligence / machine learning (AI / ML) data between a first node and a second node based on a user plane (UP) in a wireless communication system. The method includes: providing a Protocol Data Unit (PDU), wherein the PDU includes at least one of the following: a Quality of Service (QoS) Flow Identifier (QFI) indicating a Quality of Service (QoS) flow carrying AI data information; and / or a Target Identifier (ID) for routing the AI ​​data information to a node, wherein a User Equipment (UE) transmits the Target ID to an Access and Mobility Management Function (AMF) via a PDU Session Establishment Request message, the AMF sends an Information Element (IE) to a Next Generation Base Station (gNB) via an N2 PDU Session Request, and forwards the IE to a Service Management Function (SMF) via an N4 Session Establishment, and / or forwards the IE to a User Plane Function (UPF) via an N4 Session Modification, to establish routing termination node information between the first node, the gNB, and the UPF.

[0016] In a fourth aspect of this disclosure, a method for data management based on a UP in a wireless communication system is provided to perform the above method.

[0017] In a fifth aspect of this disclosure, a network is provided, including a memory, a transceiver, and a processor connected to the memory and the transceiver. The processor is configured to perform the methods described above.

[0018] In a sixth aspect of this disclosure, a user equipment is provided, including a memory, a transceiver, and a processor connected to the memory and the transceiver. The processor is configured to perform the methods described above.

[0019] In a seventh aspect of this disclosure, a wireless communication system is provided, including a memory, a transceiver, and a processor connected to the memory and the transceiver. The processor is configured to perform the methods described above.

[0020] In an eighth aspect of this disclosure, a non-volatile computer-readable storage medium is provided having instructions stored thereon that, when executed by a computer, cause the computer to perform the methods described above.

[0021] In a ninth aspect of this disclosure, a chip is provided, including a processor configured to invoke and run a computer program stored in a memory to cause a device on which the chip is mounted to perform the methods described above.

[0022] In a tenth aspect of this disclosure, a computer-readable storage medium is provided, wherein a computer program is stored that enables a computer to perform the methods described above.

[0023] In the eleventh aspect of this disclosure, a computer program product is provided, including a computer program that enables a computer to perform the methods described above.

[0024] In a twelfth aspect of this disclosure, a computer program is provided that enables a computer to perform the above-described methods. Attached Figure Description

[0025] To more clearly illustrate the embodiments or related technologies of this disclosure, the illustrations in the embodiments will be briefly described below. Obviously, these illustrations are only some embodiments of this disclosure, and those skilled in the art can obtain other illustrations without incurring additional costs based on these illustrations.

[0026] Figure 1 This is a schematic diagram of the overall AI framework for wireless communication in existing technologies.

[0027] Figure 2 shows the direction of AI data transmission / transmission.

[0028] Figure 3 This is a schematic diagram of the user plane protocol stack.

[0029] Figure 4 A schematic diagram of the classification and UP marking of Quality of Service (QoS) flows and their mapping relationship with access network resources.

[0030] Figure 5A This is a schematic diagram of the ADAP sublayer structure in an embodiment of this disclosure.

[0031] Figure 5B This is a schematic diagram of the ADAP sublayer structure in another embodiment of this disclosure.

[0032] Figure 6 This is a schematic diagram of the ADAP data PDU format in an embodiment of this disclosure.

[0033] Figure 7 This is a schematic diagram illustrating the possible mapping relationship between ADFI and AI data stream features in an embodiment of this disclosure.

[0034] Figure 8 This is a schematic diagram illustrating potential AI data transmission in a UP according to an embodiment of this disclosure.

[0035] Figure 9 This is a schematic diagram illustrating potential AI data transmission in a UP according to an embodiment of this disclosure.

[0036] Figure 10 This is a schematic diagram illustrating potential AI data transmission in a UP according to an embodiment of this disclosure.

[0037] Figure 11 This is a schematic diagram illustrating potential AI data transmission in a UP according to an embodiment of this disclosure.

[0038] Figure 12 This is a schematic diagram illustrating potential AI data transmission in a UP according to an embodiment of this disclosure.

[0039] Figure 13 This is a schematic diagram illustrating potential AI data transmission in a UP according to an embodiment of this disclosure.

[0040] Figure 14A This is a schematic diagram of AI data transmission in a PDU session according to an embodiment of the present disclosure.

[0041] Figure 14B This is a schematic diagram of AI data transmission in a PDU session according to another embodiment of the present disclosure.

[0042] Figure 15 This is a schematic diagram of the AI ​​data stream UP protocol stack according to an embodiment of the present disclosure.

[0043] Figure 16 This is a block diagram of one or more UEs and a network / gNB in ​​a communication network system according to an embodiment of this disclosure.

[0044] Figure 17 This is a block diagram of a wireless communication system according to an embodiment of the present disclosure. Detailed Implementation

[0045] The embodiments of this disclosure will be described in detail with reference to the accompanying drawings, outlining the technical problems, structural features, objectives, and effects. Specifically, the terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0046] This invention relates to wireless communication systems in which AI / ML algorithms are implemented through a Radio Access Network (RAN). Specifically, it relates to the new AI / ML-related SIDs for the Non-Line-of-Sight Air Interface (NRAir Interface) in Rel.18, a standard approved at 3GPP RAN plenary meeting 94e in December 2022, with related discussions initiated by RAN1 in May 2022. This invention provides a method for transmitting AI / ML data between a UE and network nodes (gNB, LMF, etc.) via a UP (Uplink Provider). Specifically, this method achieves AI data transmission through a UP-based scheme, significantly reducing signaling overhead and transmission conflicts compared to traditional CP-based wireless communication methods. Invention Effects

[0047] We provide a UP-based AI data transmission / transmission method, which has not been discussed in current research; it supports the transmission of large amounts of AI data at once; and it reduces overhead and collisions compared to the CP method in wireless communication.

[0048] Based on the background and introduction, the following problems exist in implementing the UP-based method for AI data information transfer / transmission between UE and NW. Question 1: How to identify AI data information.

[0049] Service data is transmitted using a UP-based method, and its UP protocol stack is as follows: Figure 3 As shown. This means that the terminal nodes are UE and UPF, while the gNB only transmits service data transparently. In 5G non-radio networks, service data is encapsulated in IP packets and mapped to the CN's QoS flow to distinguish different QoS "flows" in the PDU session. The upper-layer protocols of the CN and UE use QoS Flow Identifiers (QFIs) to mark QoS flows for transmission. In the Access Network (AN), SDAP maps QoS flows to the DRB through mapping rules and marks them with QFIs. However, the DRB is only used to carry service data and cannot decode its content. For AI data acquisition, the terminal node not only needs to identify the AI ​​data, but also needs to understand its content to enable and process related functions. Traditional UP-based methods do not cover the case where the AI ​​data acquisition terminal node is a gNB, so the gNB cannot obtain the data content through traditional methods. In addition, QoS characteristics include parameters such as resource type, priority, packet loss rate budget, and bit error rate, which are irrelevant to some content in the AI ​​data (such as location information, L3 cell / beam measurement results, L1 channel state information measurement results, etc.). Therefore, it is necessary to clarify how terminal nodes (UE, gNB, CN functions) identify AI data and its content. Question 2: How to route AI data information.

[0050] As mentioned and discussed in the introduction and Question 1, 5G CN and 5G RAN map IP packets to QoS and DRB to ensure quality of service. Figure 4 As shown, this process involves two steps: first, mapping IP packets to QoS flows (non-access stratum, NAS), and then mapping them from QoS to DRBs (access stratum, AS).

[0051] Figure 16 This document illustrates, in some embodiments, communication between multiple UEs 10 and a network / gNB 20, constituting a communication network system 40. The communication network system 40 includes one or more UEs 10 and a network / gNB 20. Each UE 10 may include a memory 12, a transceiver 13, and a processor 11 connected to the memory 12 and transceiver 13. The network / gNB 20 may include a memory 22, a transceiver 23, and a processor 21 connected to the memory 22 and transceiver 23. The processor 11 or 21 may be configured to implement the functions, procedures, and / or methods set forth in this specification. Various layers of the wireless interface protocol may be implemented in the processor 11 or 21. The memory 12 or 22 is functionally connected to the processor 11 or 21 and stores various information for use by the processor 11 or 21. The transceiver 13 or 23 is also functionally connected to the processor 11 or 21 and is responsible for transmitting and / or receiving wireless signals.

[0052] Processor 11 or 21 may include application-specific integrated circuits (ASICs), other chipsets, logic circuits, and / or data processing devices. Memory 12 or 22 may include read-only memory (ROM), random access memory (RAM), flash memory, memory cards, storage media, and / or other storage devices. Transceiver 13 or 23 may include baseband circuitry for processing radio frequency signals. When the implementation is in software form, the techniques described herein can be implemented by modules (e.g., programs, functions, etc.) that perform the functions described herein. These modules may be stored in memory 12 or 22 and executed by processor 11 or 21. Memory 12 or 22 may be integrated inside processor 11 or 21 or located outside processor 11 or 21, in which case the memory may be communicatively connected to processor 11 or 21 in various ways known in the art. Overall Solution How to identify whether a data packet is AI data.

[0053] The basic principle for identifying whether a data packet is AI data or traditional service data is achieved through a newly defined adaptation layer or a newly defined instruction. (1) For the newly defined adaptation layer: Design an adaptation layer for AI data transmission, define its supported functions, and configure the information in the new adaptation layer header. (2) Regarding the newly defined instructions: New instructions are defined to distinguish information in AI data, and these instructions are also associated with some characteristics of AI data streams.

[0054] The two solutions mentioned above can be found in Example 1. How to solve the problem of AI data transmission.

[0055] AI data can be transmitted between the UE and gNB, and between the UE and CN. In the case of AI data transmission between the UE and gNB, routing of AI data between the UE and gNB is achieved through configuration information such as ADFI, mapping rules, and DRB. The implementation methods are shown in Examples 2, 3, 4, and 5. For the case of AI data transmission between the UE and CN, routing of AI data between the UE and CN is achieved by configuring DRB and target ID in the Protocol Data Unit (PDU) session. The implementation method is shown in Example 6.

[0056] In this disclosure, certain concepts and definitions are defined as follows: AI Data Information: Information related to AI-based methods, including but not limited to at least one of the following: AI data content, AI-specific use cases, and AI-specific functions / actions, wherein: (1) AI Data Content: refers to data used for AI functions / operations. Potential AI data content may include: L1-RSRP measurements and / or SSBRI / CRI (beam ID) of a set of beams in one or more time instances; and / or CSI-RS measurements (target H or target precoder for model training / monitoring); and / or channel impulse response (CIR) information; and / or cell ID, and / or area ID, and / or carrier frequency, and / or UE / network antenna beam configuration / ID, and / or timestamp, and / or UE location / movement / rotation, and / or measurement accuracy, and / or measurement resolution, and / or UE location (real label, available from LMF), and / or prediction accuracy, and / or intermediate key performance indicators (KPIs), such as SGCS; and / or monitoring based on traditional CSI, and / or additional traditional CSI reports, etc. In addition, it also indicates some AI model information, such as AI model structure, AI model parameters, AI model layers, etc. (2) AI-specific use cases: refers to use cases / sub-use cases in wireless systems that use AI-based methods, such as CSI feedback enhancement, beam management, and improved positioning accuracy. (3) AI-specific functions / operations: refers to certain aspects defined in lifecycle management (LCM), such as model training, model inference, model monitoring, model update, model switching, model selection, etc. AI data stream: refers to expressing AI data information through a virtual stream, which is called AI data stream; and is identified by a new indicator (ADFI). In addition, its characteristics may include priority, latency limit, data size, etc.

[0057] To avoid confusion, the data shown in this patent represents only AI data collected through a data collection framework, which may be logical entities and / or functional entities.

[0058] In some embodiments, a method for transmitting AI / ML data between a first node and a second node based on UP in a wireless communication system includes: receiving AI data information and / or converting AI data information into an AI data stream through an ADAP sublayer; the AI ​​data information is forced into an AI data stream according to its characteristics; the ADAP sublayer maps the AI ​​data stream into a QoS stream and / or RB (SRB and / or DRB), and / or RLC channel / bearer, and / or logical channel.

[0059] In some embodiments, the ADAP sublayer includes at least one of the following: AI Data Stream Identifier (ADFI); Quality of Service Flow Identifier (QFI); Source ID; Target ID; Serial number (SN); D / C indication; R indicates.

[0060] In some embodiments, the configuration of ADAP includes: ADFI, QFI), mapping rules for AI data information, source ID, target ID, default radio bearer and / or serial number, and the configuration of ADAP is performed by RRC signaling and / or RB.

[0061] In some embodiments, the ADAP sublayer further provides at least one of the following functions: transmitting AI data information, indicating / distinguishing AI data information, segmenting AI data information, mapping AI data information to QoS streams, and / or radio bearers (DRB, RLC channel, and / or logical channel).

[0062] In some embodiments, the ADAP sublayer sits above the Service Data Adaptation Layer (SDAP), Packet Data Convergence Layer (PDCP), RLC Layer, and / or Media Access Control Layer (MAC).

[0063] In some embodiments, the receiving node distinguishes data packets identified as AI data information from traditional service data packets using one or more of the following methods: QFI; RB; Instructions; and / or Target ID.

[0064] In some embodiments, a method for transmitting AI / ML data between a first node and a second node based on UP in a wireless communication system includes: AI data information / AI data streams are indicated / differentiated via ADFI; and / or The QoS flow carrying AI data is indicated by QFI to disguise the AI ​​data flow as a traditional QoS flow.

[0065] In some embodiments, ADFI is associated with at least one of the following AI data stream information: priority level, latency limit, and data size; wherein the value of ADFI, AI data stream characteristics, and the mapping between ADFI and AI data stream characteristics are fixed, pre-configured, or configured, and the configuration of these values ​​is performed by Network Access Service (NAS), RRC messages, Media Access Control Control Element (MAC CE), Uplink Control Information (UCI), and / or Downlink Control Information (DCI).

[0066] In some embodiments, AI data information comes from any layer of the L2 layer and is transmitted through the SDAP layer, PDCP layer, RLC layer, MAC layer and PHY layer, or through the MAC layer and PHY layer.

[0067] In some embodiments, SDAP configured by RRC, NAS, UCI, or DCI provides the following functionality: Transmitting AI data information: AI data information is enforced / driven in the AI ​​data stream, and the AI ​​data stream represents the finest granularity of the differentiation of AI data information; AI data information is indicated / differentiated via QFI, indicator, or ADFI; and / or Mapping AI data to RB, where SDAP configuration includes: ADFI, QFI, indications for AI data streams, source ID, destination ID, mapping rules for AI data information, and / or default radio bearer.

[0068] In some embodiments, the MAC configured by RRC, NAS, UCI, or DCI provides the following functionality: Transmitting AI data: AI data is enforced / driven within the AI ​​data stream, which represents the finest granularity of differentiation within the AI ​​data; and / or AI data information is indicated / distinguished through QFI, indication, or ADFI, where the MAC configuration includes: ADFI, indication for AI data stream, target ID, and / or mapping rules for AI data information.

[0069] In some embodiments, a method for transmitting AI / ML data between a first node and a second node based on a UP in a wireless communication system includes: providing a PDU, wherein the PDU includes at least one of the following: a QFI indicating a QoS stream carrying AI data information; and / or This is used to route AI data information to the target ID of the node. The UE sends the target ID to the AMF through the PDU session establishment request message. The AMF sends the information element IE to the gNB through the N2 PDU session request, and forwards the IE to the service management function SMF through the N4 session establishment, and / or forwards the IE to the UPF through the N4 session modification, so as to establish the routing termination node information between the UE, gNB and UPF.

[0070] In some embodiments, the method for transmitting AI / ML data between a first node and a second node based on UP in a wireless communication system further includes: configuring a threshold for selecting a CP-based method or a UP-based method, wherein the threshold is configured, pre-configured, or fixed by NAS, RRC messages, MAC CE, UCI, and / or DCI based on data size, delay, and / or data type.

[0071] In some embodiments, the AI ​​data information is transmitted using either a CP-based method or a UP-based method, depending on the type or situation in which the AI ​​data information is used.

[0072] In some embodiments, the first node or the second node is one of the UE, gNB, or CN functions.

[0073] In some embodiments, a UP-based method is used to perform the above-described method for data management of a wireless communication system.

[0074] In some embodiments, the network includes: a memory; a transceiver; and a processor connected to the memory and the transceiver; wherein the processor is used to perform any of the methods described above.

[0075] In some embodiments, the user equipment includes: a memory; a transceiver; and a processor connected to the memory and the transceiver; wherein the processor is configured to perform any of the methods described above.

[0076] In some embodiments, instructions are stored on a non-volatile machine-readable storage medium that, when executed by a computer, cause the computer to perform any of the methods described above.

[0077] In some embodiments, a chip includes a processor configured to invoke and execute a computer program stored in a memory to enable a device on which the chip is mounted to perform any of the methods described above.

[0078] In some embodiments, a computer-readable storage medium is provided, wherein a computer program is stored that enables a computer to perform any of the methods described above.

[0079] In some embodiments, the computer program product includes a computer program that enables a computer to perform any of the methods described above.

[0080] To reduce the amount of AI data transmitted via UP-based methods, this patent disclosure proposes the following solutions, which reduce signaling overhead and transmission conflicts in wireless communication compared to CP-based methods. Example 1: Identifying AI data and its related information.

[0081] In this embodiment, two options are provided to solve problem 1 above, as follows: Option 1: Identify AI data information between UE and gNB

[0082] When AI data is transmitted between the UE and gNB, the following alternative solutions can be considered to enable the nodes (UE, gNB) to recognize the AI ​​data: Alternative Solution 1: Use ADAP to identify AI data information.

[0083] A new sublayer called ADAP is introduced, which aims to solve the problem of AI data transmission between the UE and gNB. Through ADAP, the two end nodes (UE and gNB) can identify and transmit AI data. Figure 5 illustrates a possible structure of the ADAP sublayer, but does not limit its implementation.

[0084] In Figure 5(a), AI data is forcibly converted into an AI data stream through the ADAP sublayer. ADAP receives the AI ​​data and exports it as an AI data stream. In Figure 5(b), AI data is forcibly converted into an AI data stream based on its characteristics. ADAP receives the AI ​​data stream and exports it as a disguised QoS stream, and / or RB (SRB, DRB), and / or RLC channel / bearer, and / or logical channel, etc. It is worth noting that AI data (obtained through a specific AI acquisition method) can be stored in the ADAP sublayer and / or in other layers of NR layer 2.

[0085] Furthermore, Figure 5 is based on the radio interface protocol architecture in NR. More specifically, it can reside above the SDAP sublayer, PDCP layer, RLC layer, or MAC layer. The ADAP entity resides in the ADAP sublayer, and one or more ADAP entities can be defined for each UE. Each ADAP entity has a transmit part and a receive part.

[0086] The above discussion shows that: The ADAP sublayer supports at least one of the following functions: ● Transmitting AI Data: Used to transmit / deliver AI data, which can be enforced / driven within the AI ​​data stream. The AI ​​data stream represents the finest granularity of AI data differentiation; or ● AI Data Information Indication / Distinction: Used to distinguish AI data information / AI data streams; each AI data stream is classified and labeled using ADFI; or ● Segmentation of AI data information indicates support for segmentation of AI data information / AI data streams when the amount of transmitted AI data information is large; or ● Disguising AI data information / AI data streams as QoS streams and / or radio bearers indicates support for disguising AI data information as certain traditional concepts, such as QoS streams and / or radio bearers (DRB, SRB) and / or RLC channels / bearers, and / or logical channels; or ● Mapping AI data information to QoS streams and / or radio bearers (DRB, RLC channel, LCID) depends on the location of the ADAP layer in the L2 architecture.

[0087] ADAP configuration is used to set configurable ADAP parameters for ADAP entities, and includes at least one of the following: ●ADFI: Used to uniquely identify AI data streams; or ●QFI: Used to disguise AI data / AI data streams as QoS streams; or ● Mapping rules for AI data information: These rules aim to map AI data information to RBs, where RBs can be DRBs / SRBs; or RLC channels / bearers; or logical channels; or, if the AI ​​data information can be driven into an AI data stream, and ADFI can identify that AI data stream, then a series of ADFIs mapping to RBs are indicated; or ● Source ID and Target ID: The source ID indicates the source of the AI ​​data collection; the target ID indicates the destination of the AI ​​data transmission / delivery. AI data collection is based on UE, cell, scenario / use case, region, site, function, model, etc. Therefore, the source ID and target ID can be UE ID, cell ID, scenario / use case ID, region ID, site ID, function ID, and model ID; or... ● Default Radio Bearer: Indicates whether this radio bearer is the default RB for AI data information. This field is set to true in at most one ADAP configuration instance and false in all other instances; or ●SN: A sequence number used to indicate / identify large data segments; its length can be predefined and / or configured.

[0088] ADAP is configured via RRC signaling, such as RRC Reconfiguration and / or RRC Resume. More specifically, ADAP configuration can be contained within an RRC; and / or it can be contained within an RB, which is contained within the RRC signaling; additionally, the RB can be a newly defined bearer or a legacy bearer.

[0089] ADAP PDU is a byte-aligned bit string (i.e., a multiple of 8 bits in length). Figure 6 The format of BAP data PDUs is shown.

[0090] Among them, parameters such as SN, ADF I, QFI, source ID, and target ID are as described above, and AI data information is described in the overall solution; other parameters are explained as follows: ●D / C: The D / C bit indicates whether the ADAP PDU is an ADAP data PDU or an ADAP control PDU; ●R: Reserved bit. In this version of the specification, the reserved bit should be set to 0. The receiver should ignore the reserved bit. Alt2: AI data information is identified based on ADFI.

[0091] ADFI (Advanced Graphical Functionality) is used to label AI data streams to address the transmission of AI data between the UE and gNB. With ADFI, the two terminal nodes (UE and gNB) can identify AI data. Similar to Option 1, AI data is forced / driven within the AI ​​data stream, which represents the finest granularity of AI data differentiation. Simultaneously, AI data can be collected using AI-specific methods (e.g., L1 measurement, L3 measurement, MDT, etc.) and stored in other layers of NR Layer 2.

[0092] In addition, ADFI is associated with some information about AI data streams, which may include at least the following: action - Priority: Indicates the priority of AI data resource scheduling. Typically, this depends on the AI-specific application scenario and / or AI-specific functions / behaviors. For example, AI data resources for online model updates have a higher priority than those for offline model training; or - Delay Constraint: Indicates the maximum / minimum delay constraint for AI data transmission / delivery; or - Data size: Indicates the maximum size / quantity of the AI ​​data being transferred.

[0093] Figure 7 This demonstrates a possible mapping relationship between ADFI and AI data stream characteristics, defined as a one-to-one correspondence. Furthermore, the ADFI value, AI data stream characteristics, and the mapping between ADFI and AI data stream characteristics can all be fixed, pre-configured, or dynamically configured. This information can be used by NAS, RRC messages, MAC control elements, UCI, DCI, etc. In other words, nodes can identify AI data information, its content, and related information through ADFI.

[0094] Figure 7 Examples of AI data information demonstrate the uses of AI data information, such as its application scenarios / contexts and / or AI-specific functions / behaviors.

[0095] Based on the above discussion, ADFI supports at least one of the following functions: AI data information indication / distinction: Used to distinguish AI data information / AI data streams. Each AI data stream is classified and labeled using ADFI.

[0096] In some instances, ADFI can also be viewed as a newly defined QFI, and the newly defined QFI is intended to set a new value for the QFI. Option 2: AI data information recognition between UE and CN.

[0097] The newly defined QFIs are used to identify "QoS flows," which actually carry AI data rather than traditional service data. More specifically, existing QFI values ​​(ranging from 0 to 63) are expanded; for example, QFIs numbered 64 to 72 are used to identify these new QFIs to facilitate the labeling of AI data.

[0098] In 5G CN, QFI will be sent to UE, gNB and UPF respectively. The messages may include PDU session resource modification request, PDU session resource establishment request, PDU session establishment acceptance, PDU session modification command, N4 session establishment and N4 session modification, etc.

[0099] Within the RAN, the SDAP configuration should also include a newly defined QFI, which is configured by RRC messages, such as the RRCReconfiguration message. Example 2: Routing of AI data transmission / transfer between UE and gNB

[0100] This example demonstrates a potential routing of AI data information between the UE and gNB based on ADAP.

[0101] Figure 8 This diagram illustrates data routing within the UP stack. The ADAP sublayer sits above the SDAP layer. Blue lines represent traditional QoS flows, while red lines represent AI data flows. It demonstrates the transmission of AI data flows between the UE and gNB via ADAP, SDAP, PDCP, RLC, MAC, and PHY (not shown in the diagram). Figure 8 (as shown in the image). ADAP configuration.

[0102] To ensure the transmission / transmission of AI data, the following configurations should be defined first.

[0103] ADAP configuration is used to set configurable parameters for ADAP entities and should include at least one of the following: ADFI; and / or QFI; and / or default DRB; and / or mapping rules for AI data information; and / or indication of masquerading QoS flows; and / or target node ID. This information can be configured via RRC messages or is a fixed parameter. The definitions of the mapping rules for ADFI, QFI, and AI data information are the same as in Example 1, while the explanations of the other elements are as follows: ● AI data information indicator: Used to indicate that a QoS flow is a disguised Quality of Service flow; in other words, through this indicator, a QoS flow can be identified as AI data information; or ●Transmission / transfer of AI data information.

[0104] Based on the discussion in Implementation Opt1, when AI data information needs to be transmitted, the ADAP entity sending the data should: - Map AI data streams to QoS streams. In other words, AI data information will be processed at the ADAP layer, thus enabling the AI ​​data stream to be disguised as a QoS stream. - Construct ADAP data PDUs by adding ADAP headers. Submit the ADAP PDU to the DRB.

[0105] More specifically, multiple AI data streams can be mapped to a single, disguised QoS stream. ADAP can tag ADFI to identify AI data streams; this tag is used to distinguish different AI data streams. - ADFI is optional if the AI ​​data stream and the masqueraded QoS stream are mapped one-to-one, or - If there is a many-to-one (M-to-1) mapping relationship between AI data streams and masquerading QoS streams, then ADFI needs to be introduced, where M represents multiple AI data streams.

[0106] Subsequently, ADAP can use QFI to tag the AI ​​data stream, thereby disguising the AI ​​data stream as a "QoS stream" and transmitting it to the SDAP layer. Since traditional QoS streams are processed between the UE and the data network (DN), the receiving node should be able to identify whether the data packet is an AI data packet or a traditional service packet; therefore, the following solutions can be considered. ● Instead of using existing QFI values ​​(0-63), new QFIs should be used to identify spoofed QoS flows. Therefore, the number of QFIs should be expanded, and new QFIs should be used to identify spoofed QoS flows. For example, these new QFIs can be designated using QFI numbers 64-72. Furthermore, QFI numbering can be configured via RRC messages, meaning that the QFI for a spoofed QoS flow can be temporary and / or fixed. ● Use a dedicated DRB to carry spoofed QoS flows; the RRC message will negotiate which RBs can be used to carry spoofed QoS flows, or / and define only which RBs can be used to carry spoofed QoS flows. ● Use metrics in the ADAP sublayer: This metric can indicate that a "QoS flow" is a disguised QoS flow. ● Use Target ID: This metric can indicate which endpoint a "QoS flow" should be forwarded to, and can implicitly indicate that a "QoS flow" is a spoofed QoS flow.

[0107] After completing the transmission processing at the ADAP layer, the SDAP layer maps the masqueraded QoS flow to the DRB, as detailed below: If no AI data flow rules exist, the spoofed QoS flow can be mapped to the default DRB; Based on AI data flow rules, spoofed QoS flows can be mapped to dedicated radio bearers.

[0108] AI data flow rules can be configuration-based (e.g., RRC configuration) and / or fixed rules. Furthermore, masquerading QoS flows can be mapped to the same radio bearer as the QoS flow. This radio bearer can be newly defined and / or conventional, and can include DRB, and / or RLC channels, and / or LCID. Subsequently, the masquerading QoS flow is processed by PDCP, RLC, MAC, and PHY, and then sent to the receiving node using allocated AI data transmission resources. PDCP header compression / decompression is optional for AI data, and AI data can be transmitted in RLC using unacknowledged (UM) or acknowledged (AM) mode. When the terminal node receives an AI data packet sent by the sending node, it decodes the packet to obtain the AI ​​data information. Furthermore, the receiving node can distribute the received AI data to various AI function modules deployed on the receiving node based on its implementation and / or reconfiguration information (e.g., AI function / action information or latency information). Implementation Method 3: Routing for AI data transmission / transfer between UE and gNB.

[0109] This implementation demonstrates some possibilities regarding the potential routing of AI data information. As discussed in Implementation 2, the transmission / transmission of AI data information can be based on the ADAP sublayer. In addition to Implementation 2, the ADAP sublayer can also be located above PDCP, or RLC, or MAC.

[0110] In some instances, the corresponding data routing in the UP stack is as follows: Figure 9 As shown, the ADAP sublayer sits above the PDCP layer. The blue line represents traditional QoS flows, and the red line represents AI data flows. This describes the data flow through ADAP, PDCP, RLC, MAC, and PHY (in...). Figure 9 AI data stream transmission between the UE and gNB (not shown in the image).

[0111] In some instances, the corresponding data routing in the UP stack is as follows: Figure 10 As shown, the ADAP sublayer sits above the RLC layer. The blue line represents the traditional QoS flow, and the red line represents the AI ​​data flow, describing the flow through ADAP, RLC, MAC, and PHY (in...). Figure 10 AI data stream transmission between the UE and gNB (not shown in the image).

[0112] In some instances, the corresponding data routing in the UP stack is as follows: Figure 11 As shown, the ADAP sublayer is located above the MAC layer. The blue line represents the traditional QoS flow, and the red line represents the AI ​​data flow, describing the flow through ADAP, MAC, and PHY (in...). Figure 11 AI data stream transmission between the UE and gNB (not shown in the image). ADAP Configuration

[0113] Regarding the transmission / transmission of AI data information, the following configurations should be clarified first.

[0114] ADAP configuration is used to set configurable parameters for ADAP entities, which include at least one of the following: ADFI; and / or default DRB; and / or mapping rules for AI data information; and / or indications for AI data information; and / or target node ID. These parameters can be configured via RRC messages and / or are fixed. The definitions of ADFI, default DRB, mapping rules for AI data information, indications for AI data information, and / or target ID are the same as in Example 2. AI data and information transmission.

[0115] Based on the discussion in Implementation Opt1, when AI data information needs to be transmitted, the ADAP entity sending the data should: - Map the AI ​​data stream to an RB, which includes a default DRB and / or RLC channels / bearers, and / or logical channels; - Construct the ADAP PDU by adding an ADAP header; Submit this ADAP PDU to the lower level.

[0116] More specifically, one or more AI data streams can be mapped to RBs. ADAP can tag ADFIs to identify AI data streams; this tag is used to distinguish different AI data streams. - ADFI is optional if the AI ​​data stream is mapped one-to-one with the RB; or - If the AI ​​data stream and RB are many-to-one mapped, then ADFI is required, where M represents multiple AI data streams.

[0117] The receiving node should be able to distinguish whether the transmitted data packets are AI data. The following solutions can be considered: 1) Use an indicator in the ADAP layer to indicate that the transmitted data packet is AI data; or 2) Use a dedicated RB to carry AI data; RRC message negotiation can be used for the RB that carries AI data information, and / or fix the RB that can be used to carry AI data information.

[0118] After the ADAP layer process is completed, the AI ​​data is mapped to an appropriate Resource Block (RB). Alternatively, the AI ​​data stream and the QoS stream can also be mapped to the same RB. This RB can be newly defined and / or a legacy RB. Subsequently, the AI ​​data stream undergoes processing by PDCP, RLC, MAC, and PHY, and is then sent to the receiving node using the allocated AI data transmission resources. Compression / decompression of the PDCP header is optional for AI data, and AI data can be transmitted via RLC using UM or AM. When the receiving node receives an AI data packet from the sending node, it decodes the packet to obtain the AI ​​data. Furthermore, based on its implementation and / or reconfiguration information (such as AI function / action information or latency information), the receiving node can distribute the received AI data to various AI function modules deployed on it. Example 4: AI data transmission / delivery routing between UE and gNB

[0119] This example demonstrates a potential route based on AI data information between the UE and gNB using ADFI.

[0120] Figure 12 The diagram illustrates data routing within the UP stack. Dashed lines represent traditional QoS flows, while chained lines represent AI data flows. It demonstrates the AI ​​data flow transfer between the UE and gNB via SDAP, PDCP, RLC, MAC, and PHY (this process is not shown in the diagram). Figure 12 (details described in the text). Configuration.

[0121] SDAP provides at least one of the following functions: ● Transmitting AI Data: Used to transmit / deliver AI data, which can be forcibly applied / driven within the AI ​​data stream. The AI ​​data stream represents the finest granularity of AI data differentiation; or ● Indicators / Differentiation of AI Data Information: Used to differentiate AI data information / AI data streams. This differentiation can be achieved through the new QFI, or by using new indicators or ADFI; or ●Mapping AI data information to RB: AI data information / AI data streams can be mapped to DRB.

[0122] To achieve the above functions, SDAP can be configured via RRC, NAS, UCI, or DCI, and must include at least one of the following information: ● ADFI: Used to uniquely identify AI data streams; or ● QFI: The new QFI indicates that the data stream is an AI data stream; or ● An indicator that specifies an AI data stream, signifying that the data stream is an AI data stream; or an ID identifier that indicates the location of AI data transmission / delivery. AI data collection can be based on UE, cell, scenario / use case, region, site, function, or model. Therefore, the source ID and target ID can be UE ID, cell ID, scenario / use case ID, region ID, site ID, function ID, or model ID; or ● AI Data Mapping Rules: These are the rules for mapping AI data to Radio Bearers (RBs). The radio bearer can be a DRB / SRB; an RLC channel; a logical channel; or, if the AI ​​data can be driven as an AI data stream and the ADFI can recognize it, then it represents the ADFI list of AI data streams mapped to the radio bearer; or... ● Default Wireless Bearer: Indicates whether this is the default wireless bearer for AI data. This field should be set to true in at most one ADAP configuration instance and false in all other instances. Transmission of AI data information

[0123] Based on the discussion in Implementation Opt1, when AI data information needs to be transmitted, the sending SDAP entity should: - Mapping AI data streams to DRB: AI data information / AI data streams are processed by the SDAP layer and mapped to the DRB. Furthermore, it can also disguise AI data streams as QoS streams. - Build an SDAP PDU by adding an SDAP header. Submit the SDAP PDU to the DRB.

[0124] More specifically, single or multiple AI data streams can be directly mapped to the DRB through the SDAP layer. Therefore, SDAP can tag ADFIs to identify the AI ​​data stream, and thus distinguish between different AI data streams. - If there is a one-to-one mapping between the AI ​​data stream and the DRB, then ADFI is optional; or - If there is a many-to-one mapping between the AI ​​data stream and the DRB, then ADFI must be used, where "M" represents multiple AI data streams.

[0125] Additionally, AI data streams can also be identified via QFI. Receiving nodes should distinguish between AI data packets and traditional service data; the following solutions may be considered: ● A data stream indicator, which can be ADFI, a newly defined QFI, or other defined indicators. For example, if the received data stream carries ADFI, it indicates that it is an AI data stream; if the received data stream carries a newly defined QFI, it also indicates that it is an AI data stream. ● Use a dedicated DRB to carry spoofed QoS flows; RRC message negotiation can be used for RBs carrying spoofed QoS flows, or simply define which RBs can be used to carry spoofed QoS flows; or ●Target ID: Indicates which endpoint the data stream should be forwarded to. This method can also implicitly indicate that the data stream is AI data information.

[0126] The sent SDAP processes the AI ​​data stream and forwards it to the lower PDCP layer. Specifically, the sent SDAP maps the AI ​​data stream to the DRB, as detailed below: If the AI ​​data stream rule does not exist, the AI ​​data stream can be mapped to the default DRB; AI-based data flow rules are mapped to a dedicated DRB.

[0127] AI data stream rules can be based on configuration (e.g., RRC configuration) and / or fixed settings. More specifically, RBs that can be used to carry AI data can be negotiated using RRC messages, or such RBs can be defined. Furthermore, AI data streams and QoS streams can also be mapped to the same DRB. The RB can be newly defined and / or a traditional RB, where the RB includes a DRB, and / or an RLC channel, and / or an LCID. After the sending node maps the AI ​​data stream to the DRB, the AI ​​data stream undergoes PDCP, RLC, MAC, and PHY processing, and is then sent to the receiving node using allocated AI data transmission resources. PDCP header compression / decompression of the AI ​​data is optional, and the AI ​​data can be transmitted in UM or AM within the RLC. When the receiving node receives the AI ​​data packet sent by the sending node, it decodes the packet to obtain the AI ​​data information. Furthermore, the receiving node can allocate the received AI data to various AI function modules deployed on the receiving node based on its implementation and / or reconfiguration information (such as AI function / action information or latency information). Example 5: AI data transmission / delivery routing between UE and gNB.

[0128] This example demonstrates potential AI data routing between the UE and gNB based on ADFI.

[0129] Figure 13 This demonstrates the data routing in the user plane stack, where blue lines represent traditional QoS flows and red lines represent AI data flows. It shows the AI ​​data flows transmitted between the UE and gNB via MAC and PHY (this content is in...). Figure 13 (not shown in detail). Configuration.

[0130] MAC provides at least one of the following functions: ● Transmitting AI Data: Used to transmit / deliver AI data, which can be enforced / driven within the AI ​​data stream. The AI ​​data stream represents the finest granularity of AI data differentiation; or ● Indicator / Distinguish AI Data Information: Used to distinguish AI data information / AI data streams. The new QFI can be used to achieve the distinction function, or the new indicator or ADFI can be used to achieve this function.

[0131] To achieve the above functions, the MAC configuration can be performed via RRC, NAS, UCI, or DCI, and must include at least one of the following: ● ADFI: Used to uniquely identify the AI ​​data stream; or ● An indicator of AI data stream, indicating that the data stream is an AI data stream; or ● Target ID: Indicates which endpoint the data stream should be transmitted to. This information can also be used to implicitly indicate that the data stream is AI data. ● AI data information mapping rules: This rule is used to map AI data information to RB, which can be DRB / SRB; or RLC channel; or logical channel; on the other hand, if AI data information can be driven into AI data stream, and ADFI can identify the AI ​​data stream, it means that a series of AI data streams have been mapped to resource blocks by ADFI. AI data and information transmission.

[0132] Based on the discussion in Implementation Opt1, when AI data information needs to be transmitted, the sending MAC entity should perform the following operations: - Build a MAC PDU by adding a MAC header. Submit this MAC PDU to the PHY.

[0133] More specifically, one or more AI data streams can be mapped to logical channels, while ADFI is used to identify the AI ​​data streams. Both the mapping rules and ADFI mentioned above are configured and / or fixed. - If the AI ​​data stream and DRB have a one-to-one mapping, then ADFI is optional; or - If the AI ​​data stream and DRB are a many-to-one (M-to-1) mapping, then ADFI is required, where M represents multiple AI data streams.

[0134] The receiving node needs to distinguish between AI data packets and traditional service data packets. The following solutions can be considered: ● An indication of the data stream, which is an ADFI or other defined indication. For example, if the received data stream carries an ADFI, then the data stream is an AI data stream; ● Use a dedicated logical channel to carry AI data; RRC message negotiation can be used to negotiate the logical channel for carrying AI data, and / or define only the RB that can be used to carry the logical channel; in addition, new LCIDs can be defined or extended; or ●Target ID: This ID indicates which terminal node the data stream should be delivered to. This ID can also be used to implicitly indicate that the data stream is artificial intelligence data information.

[0135] After the sending node maps the AI ​​data stream to a logical channel, the AI ​​data stream undergoes PHY processing and is then sent to the receiving node using the allocated AI data transmission resources. When the receiving node receives the AI ​​data packets sent by the sending node, it decodes these packets to obtain the AI ​​data information. Furthermore, the receiving node can allocate the received AI data to various AI function modules deployed on the receiving node based on its implementation and / or reconfiguration information (e.g., AI function / operation information or latency information). Example 6: AI data transmission / delivery routing between UE and CN function

[0136] This embodiment demonstrates the potential routing of AI data information between the UE and CN functions (including the Learning and Management Function (LMF) and other newly defined AI-related CN functions).

[0137] 5G PDU sessions support IP, Ethernet, and unstructured service data. However, these service data differ significantly from AI data, as detailed in Question 2. Therefore, by enhancing traditional PDU sessions, PDU sessions can also support the transmission of AI data, specifically including: a) AI data and QoS streams can be transmitted within the same PDU session; b) AI data is carried in a dedicated PDU session specifically for AI data transmission, as detailed in Figure 14.

[0138] Based on the above discussion, in order to support the transmission of AI data information between the UE and the CN, the following information needs to be considered for the functions in the traditional PDU session: ● QFI: A new QFI value needs to be defined or extended to distinguish whether the QoS stream is AI data information. See Example 1 for details. ● Target ID: Used to route AI data to the correct node. More specifically, the UE carries the target ID to the AMF via a PDU session establishment request message. The AMF can send the IE to the gNB via an N2 PDU session request message, and can also forward it to the SMF. The SMF forwards the IE to the UPF via an N4 session establishment and / or N4 session modification message. If all the above steps are successful, the UE, gNB, and UPF will have the correct routing endpoint information.

[0139] Figure 15 The user plane stack of the AI ​​data stream is demonstrated, revealing the transfer of AI data streams between the UE and CN functions via PDU sessions. Alternatively, this approach can also be used for AI data transmission between the UE and gNB. Generally, the AI ​​data stream is transmitted from the UE to the UPF, and then forwarded by the UPF to the gNB via the N3 tunnel. The routing process for uplink AI data can be described as follows:

[0140] First, a PDU session supporting the transmission of AI data needs to be established. This can be a traditional session or a newly defined session. The routing process is as follows:

[0141] AI data can be encapsulated into IP packets. When the PDU layer receives an IP packet, it first distinguishes whether the IP packet is AI data or traditional service data. Specifically, the PDU layer can identify the service type based on the Differentiated Service (DS) in the IP header. For example, a new differentiated service can be defined to identify AI data. Through the differentiated service, the PDU packet can distinguish whether the IP packet is AI data or traditional service data.

[0142] In NAS, IP packets are mapped to QoS. To distinguish "QoS flows" from traditional QoS flows, a new QFI can be used to identify spoofed QoS, and this identifier should not use existing QFI values ​​(0-63). Therefore, the number of QFIs should be expanded, and the new QFI should be used to identify QoS flows used for AI data. Furthermore, "QoS flows" are associated with the target node ID because the function collecting AI data may be deployed on certain CN functions, such as Location Management Functions (LMF) or newly defined AI functions, rather than residing in the DN. The aforementioned IEs are configured during PDU session establishment / modification.

[0143] In the received access layer (AS), the QoS flow of AI data undergoes processing via SDAP, PDCP, RLC, MAC, and PHY, and is then transparently transmitted to the CN function through the gNB, utilizing the allocated AI data transmission resources. The details of the access layer processing (SDAP, PDCP, RLC, MAC, PHY) are similar to those in Example 4. Furthermore, SDAP and / or the tunnel endpoint protocol-user plane (GTP-U) includes the target node ID. Example 7: AI data transmission selection based on CP or UP.

[0144] If a node can simultaneously support the transmission / transmission of AI data via both CP-based and UP-based methods, then when transmitting / transmitting certain AI data, it is necessary to consider whether to use the CP-based or UP-based method. The following solutions have been proposed: Option 1: Define a threshold to select either a CP-based or UP-based method for transmitting AI data. This threshold can be set based on factors such as data size, latency, and data type; the threshold can be configurable, pre-configured, or fixed.

[0145] In some examples, if the threshold is based on data size, then the downlink threshold is defined as 60KB. If the size of the AI ​​data to be transmitted exceeds 60KB, the node should choose the UP-based method; if the size of the AI ​​data to be transmitted is less than 60KB, the node should choose the CP-based method. Option 2: Fix the data used for a certain type or use case to select either a CP-based or UP-based approach. For example, define AI data information for Channel State Information (CSI) and Service Management (BM) use cases to select a CP-based approach, regardless of the size of the AI ​​data information.

[0146] In Option 1, the threshold can be configured via RRC messages, NAS, MAC CE, UCI, or DCI.

[0147] Figure 17 This is a block diagram of a wireless communication system 700 according to an embodiment of the present disclosure. The embodiments described in this disclosure can be integrated into the system by any appropriately configured hardware and / or software. Figure 17 System 700 is shown, including radio frequency (RF) circuitry 710, baseband circuitry 720, application circuitry 730, memory / storage 740, display 750, camera 760, sensor 770, and input / output (I / O) interface 780. These components are interconnected at least as shown in the figure. Application circuitry 730 may include circuitry, such as, but not limited to, one or more single-core or multi-core processors. These processors may include any combination of general-purpose and special-purpose processors, such as graphics processors and application processors. The processor may be connected to memory / storage and configured to execute instructions stored in memory / storage to support various applications and / or operating systems running on the system. Although this disclosure has been described in what is considered the most practical and preferred embodiments, it should be understood that this disclosure is not limited to the disclosed embodiments and is intended to cover various arrangements made without departing from the broadest interpretation of the appended claims.

Claims

1. A method for transmitting artificial intelligence / machine learning (AI / ML) data between a first node and a second node based on the user plane (UP) in a wireless communication system, characterized in that... include: The system receives AI data information and / or converts the AI ​​data information into an AI data stream through the AI ​​Data Adaptation Protocol (ADAP) sublayer. The AI ​​data information is forcibly converted into the AI ​​data stream according to its characteristics. The ADAP sublayer then disguises the AI ​​data stream as a Quality of Service (QoS) stream and / or a Radio Bearer (RB) (Control Bearer (SRB) and / or Data Bearer (DRB)), and / or a Radio Link Control (RLC) channel / bearer, and / or a logical channel for mapping.

2. The method for transmitting AI / ML data between a first node and a second node based on UP in a wireless communication system according to claim 1, characterized in that, The ADAP sublayer includes at least one of the following information: AI data stream identifier ADFI; Quality of Service Flow Identifier (QFI); Source Identifier ID; Target ID; Serial number (SN); D / C indication; R indicates.

3. The method for transmitting AI / ML data between a first node and a second node based on UP in a wireless communication system according to claim 1, characterized in that, The configuration of ADAP includes: AI data stream identifier ADFI, quality of service stream identifier QFI, mapping rules for the AI ​​data information, source identifier ID, target ID, default radio bearer and / or sequence number. ADAP is configured through radio resource control (RRC) signaling and / or the RB.

4. The method for realizing AI / ML data transmission between a first node and a second node based on UP in a wireless communication system according to claim 1, characterized in that, The ADAP sublayer further provides at least one of the following functions: transmitting the AI ​​data information, indicating / distinguishing the AI ​​data information, segmenting the AI ​​data information, mapping the AI ​​data information to the QoS stream, and / or the radio bearer (the DRB, the RLC channel, and / or the logical channel).

5. The method for transmitting AI / ML data between a first node and a second node based on UP in a wireless communication system according to claim 1, characterized in that, The ADAP sublayer is located above the Service Data Adaptation (SDAP) layer, Packet Data Convergence Protocol (PDCP) layer, RLC layer, and / or Media Access Control (MAC) layer.

6. The method for realizing AI / ML data transmission between a first node and a second node based on UP in a wireless communication system according to claim 1, characterized in that, The receiving node distinguishes data packets from AI data packets and traditional service data packets using one or more of the following methods: Quality of Service Flow Identifier (QFI); RB; Indicators; and / or Target identifier ID.

7. A method for transmitting artificial intelligence / machine learning (AI / ML) data between a first node and a second node in a wireless communication system based on a UP (Uploader), characterized in that, include: AI data stream identification ADFI indicates / distinguishes AI data information / AI data stream; and / or The Quality of Service (QFI) indicator is used to indicate the QoS flow carrying the AI ​​data information, thereby disguising the AI ​​data flow as a traditional QoS flow.

8. The method for realizing AI / ML data transmission between a first node and a second node in a wireless communication system based on UP according to claim 7, characterized in that, The ADFI is associated with at least one of the following information of the AI ​​data stream: priority level, latency constraint, and data size; wherein the value of the ADFI, the AI ​​data stream characteristics, and the mapping between the ADFI and the AI ​​data stream characteristics are fixed, pre-configured, or configured, and the value of the ADFI, the AI ​​data stream characteristics, and the mapping between the ADFI and the AI ​​data stream characteristics are configured through Network Access Service (NAS), Radio Resource Control (RRC) messages, Media Access Control (MAC) Control Element (CE), Uplink Control Information (UCI), and / or Downlink Control Information (DCI).

9. The method for realizing AI / ML data transmission between a first node and a second node based on UP in a wireless communication system according to claim 7, characterized in that, The AI ​​data information originates from any layer in Layer 2 (L2) and is transmitted through the Service Data Adaptation (SDAP) layer, Packet Data Convergence Protocol (PDCP) layer, Radio Link Control (RLC) layer, Media Access Control (MAC) layer, and Physical PHY layer, or through the MAC layer and the PHY layer.

10. The method for realizing AI / ML data transmission between a first node and a second node based on UP in a wireless communication system according to claim 7, characterized in that, The Service Data Adaptation Protocol (SDAP), configured by Radio Resource Control (RRC) messages, Network Access Service (NAS), Uplink Control Information (UCI), or Downlink Control Information (DCI), provides the following functions: Transmitting the AI ​​data information: The AI ​​data information is enforced / driven in the AI ​​data stream, and the AI ​​data stream represents the finest granularity of the differentiation of the AI ​​data information; The AI ​​data information is indicated / distinguished through the QFI, indication, or ADFI; and / or The AI ​​data information is mapped to the radio bearer (RB), and the SDAP configuration includes: The ADFI, the QFI, the AI ​​data stream indication, the source identifier ID, the target ID, the mapping rules of the AI ​​data information, and / or the default radio bearer.

11. The method for realizing AI / ML data transmission between a first node and a second node based on UP in a wireless communication system according to claim 7, characterized in that, The Media Access Control (MAC) layer, configured by Radio Resource Control (RRC) messages, Network Access Service (NAS), Uplink Control Information (UCI), or Downlink Control Information (DCI), provides the following functions: Transmitting the AI ​​data information: The AI ​​data information is enforced / driven in the AI ​​data stream, and the AI ​​data stream represents the finest granularity of the differentiation of the AI ​​data information; and / or The AI ​​data information is indicated / distinguished through the QFI, indication, or ADFI. The configuration of the MAC layer includes: the ADFI, the indication of the AI ​​data stream, the target identifier ID, and / or the mapping rules of the AI ​​data information.

12. A method for transmitting artificial intelligence / machine learning (AI / ML) data between a first node and a second node in a wireless communication system based on the user plane (UP), characterized in that, include: Provide a Protocol Data Unit (PDU), wherein the PDU includes at least one of the following information: The Quality of Service (QoS) flow identifier (QFI) indicates the Quality of Service (QoS) flow carrying AI data information. and / or The AI ​​data information is routed to the target identifier ID of the node. The user equipment (UE) transmits the target ID to the access and mobility management function (AMF) through a PDU session establishment request message. The AMF sends the information element (IE) to the next-generation base station (gNB) through an N2 PDU session request, and forwards the IE to the service management function (SMF) through an N4 session establishment, and / or forwards the IE to the user plane function (UPF) through an N4 session modification, so as to establish routing termination node information between the UE, the gNB, and the UPF.

13. The method for transmitting AI / ML data between a first node and a second node based on UP in a wireless communication system according to any one of claims 1 to 12, characterized in that, The method further includes configuring a threshold for selecting a control plane CP-based method or an up-plane UP-based method, the threshold being configured, pre-configured, or fixed by Network Access Service (NAS), Radio Resource Control (RRC) messages, Media Access Control (MAC) Control Element (CE), Uplink Control Information (UCI), and / or Downlink Control Information (DCI) based on data size, latency, and / or data type.

14. The method for transmitting artificial intelligence / machine learning (AI / ML) data between a first node and a second node in a wireless communication system based on UP, as described in claims 1 to 12, is characterized in that... The AI ​​data information is transmitted using either the CP-based method or the UP-based method, depending on the type or situation in which the AI ​​data information is used.

15. The method for realizing AI / ML data transmission between a first node and a second node based on UP in a wireless communication system according to claims 1 to 14, characterized in that, The first node or the second node is one of the functions of user equipment (UE), next-generation base station (gNB), or core network (CN).

16. A method for managing data based on the user plane (UP) in a wireless communication system, characterized in that, Perform the method as described in any one of claims 1 to 15.

17. A network, characterized in that, include: Memory; transceiver; as well as A processor, which is connected to the memory and the transceiver; The processor is configured to perform the method as described in any one of claims 1 to 15.

18. A user equipment, characterized in that, include: Memory; transceiver; as well as A processor, which is connected to the memory and the transceiver; The processor is configured to perform the method as described in any one of claims 1 to 15.

19. A non-volatile machine-readable storage medium, characterized in that, It stores instructions thereon that, when executed by a computer, cause the computer to perform the method as described in any one of claims 1 to 15.

20. A chip, characterized in that, include: A processor for calling and running a computer program stored in memory to cause a device on which the chip is mounted to perform the method as described in any one of claims 1 to 15.

21. A computer-readable storage medium, characterized in that, The computer program is stored thereon, which causes the computer to perform the method as described in any one of claims 1 to 15.

22. A computer program product, characterized in that, Includes a computer program, wherein the computer program causes a computer to perform the method of any one of claims 1 to 15.

23. A computer program, characterized in that, The computer program causes the computer to perform the method as described in any one of claims 1 to 15.

Images