Cooperative communication for radio access network

Abstract

For AI / ML model-based operations in a wireless radio access network, different combinations of BS-UEs are used to manage signaling traffic overhead and device power consumption due to AI / ML usage. A multi-communication mode is selectively activated based on a parameter set generated as ML parameter profile and UE ML capability profile information. UE mobility and clustering are also used as triggering methods for determining related communication modes.

Description

Technical Field

[0001] Various embodiments relate generally to the field of wireless communications. Background Art

[0002] Artificial Intelligence (AI) or Machine Learning (ML) is used in many different applications and fields as it contributes to performance improvements over existing technologies. In wireless or mobile communication networks, when two or more devices communicate wirelessly, AI / ML can also be used to achieve better performance in various use cases or applications. However, there are challenges in applying AI / ML, and some of these challenges include high signaling traffic load and increased device power consumption due to AI / ML operations in wireless devices.

[0003] In the Radio Access Network (RAN) with wireless device connections, it is necessary to consider the interaction between the mobile device (UE) and the base station (BS) and other network devices (such as mobile edge computing devices (MEC) and non-terrestrial network devices (NTN)), so that AI / ML operations in the RAN can overcome the key challenges of high signaling traffic load and increased device power consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] Figure 1 A block diagram of an example wireless communication network environment for a network device (e.g., UE, AN, gNB, or eNB) according to various aspects or embodiments is shown.

[0005] Figure 2 is a diagram illustrating example AI / ML module usage in a radio access network according to one or more embodiments.

[0006] Figure 3 is a diagram illustrating federated learning operations using an AI / ML module in a radio access network according to one or more embodiments.

[0007] Figure 4 An upper table and a lower table are shown, which illustrate multiple communication modes according to one or more embodiments.

[0008] Figure 5 ML parameters for AI / ML operations according to one or more embodiments are presented.

[0009] Figure 6 A UE ML capability profile is presented in accordance with one or more embodiments.

[0010] Fig. 7A , Figure 7B , Figure 7C and Fig.7Dis a diagram illustrating various multi-communication modes for supporting AI / ML operations according to one or more embodiments.

[0011] Figure 8 is a flow chart depicting a method of operating a communication mode utilizing AI / ML operations according to one or more embodiments.

[0012] Fig. 9 is a diagram illustrating a signaling flow for AI / ML operations utilizing a communication mode according to one or more embodiments.

[0013] Fig.10 is a flow chart illustrating a method for a CRU according to one or more embodiments.

[0014] Fig.11 is a flow chart illustrating a method for a CRU according to one or more embodiments.

[0015] Fig.12 is a diagram showing a signaling flow of a mode adopting UE cluster communication according to one or more embodiments.

[0016] Fig.13 is a graph depicting a set of Gaussian distributions according to one or more embodiments.

[0017] Fig.14 is a diagram depicting AI / ML operations based on single parameter UE clustering in accordance with one or more embodiments.

[0018] Fig.15 is a diagram depicting AI / ML operations for multiple parameter UE clustering according to one or more embodiments.

[0019] Fig.16 is a diagram illustrating operations of a BS and a UE for clustering using a codebook scheme according to one or more embodiments.

[0020] Fig.17 is a flow chart describing a method of selecting a CRU for each cluster according to one or more embodiments.

[0021] Fig.18 is a diagram illustrating UE mobility using BS-BS combined training according to one or more embodiments.

[0022] Fig.19 is a diagram illustrating UE mobility using BS-BS split training in accordance with one or more embodiments.

[0023] Fig. 20 is a flow chart illustrating a method of using BS-BS combined training and / or BS-BS split training according to one or more embodiments.

[0024] Fig.21 is a flow chart illustrating a method for triggering based on UE mobility according to one or more embodiments.

[0025] Fig. 22 is a diagram illustrating global model sharing according to one or more embodiments.

[0026] Fig.23 is a diagram illustrating global model sharing according to one or more embodiments. DETAILED DESCRIPTION

[0027] The present disclosure will now be described with reference to the accompanying drawings, wherein throughout the text, the same reference numerals are used to refer to the same elements, and wherein the structures and devices shown are not necessarily drawn to scale. The same reference numerals may be used to identify the same or similar elements in different figures. In the following description, for the purpose of explanation rather than limitation, specific details such as specific structures, architectures, interfaces, technologies, etc. are described to provide a thorough understanding of various aspects of various embodiments. However, it will be apparent to those skilled in the art who benefit from the present disclosure that various aspects of various embodiments can be practiced in other examples that deviate from these specific details. In some instances, descriptions of well-known devices, circuits, and methods are omitted to avoid blurring the description of various embodiments with unnecessary details. The embodiments herein may relate to RAN1, RAN2, 5G, etc.

[0028] As used herein, the terms "component," "system," "interface," and the like are intended to refer to computer-related entities, hardware, software (e.g., in execution), and / or firmware. For example, a component can be a processor, a process running on a processor, a controller, an object, an executable program, a program, a storage device, and / or a computer having a processing device. By way of illustration, an application running on a server and the server can also be a component. One or more components can reside within a process, and components can be located on one computer and / or distributed between two or more computers. A group of elements or a group of other components can be described herein, where the term "group" can be interpreted as "one or more."

[0029] Further, for example, the components may be executed from various computer-readable storage media having various data structures stored thereon, such as using modules. The components may communicate via local and / or remote processes, such as according to signals having one or more data packets (e.g., data from one component interacts with another component in a local system, a distributed system, and / or other systems across a network (e.g., the Internet, a local area network, a wide area network, or the like) via the signal).

[0030] As another example, a component may be a device having a specific functionality provided by a mechanical part operated by an electrical or electronic circuit system, wherein the electrical or electronic circuit system may be operated by a software application or by a firmware application executed by one or more processors. The one or more processors may be internal or external to the device and may execute at least a portion of the software or firmware application. As yet another example, a component may be a device providing a specific functionality by an electronic component without a mechanical part; the electronic component may include one or more processors therein to execute the software and / or firmware that at least partially imparts the functionality of the electronic component.

[0031] The use of the word "exemplary" is intended to present concepts in a concrete manner. As used in this application, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X employs A or B" is intended to mean any natural inclusive arrangement. That is, if X employs A; X employs B; or X employs both A and B, then in any of the aforementioned instances, "X employs A or B" is satisfied. In addition, unless otherwise specified or clear from the context to refer to a singular form, the articles "a / an" as used in this application and the appended claims should generally be interpreted to mean "one or more". In addition, if the terms "including / includes", "having / has", "with" or variations thereof are used in the detailed description and claims, these terms are intended to be inclusive in a manner similar to the term "comprising".

[0032] Mobile communications have evolved from early voice systems to highly complex integrated communication systems or platforms. It is expected that the next generation of wireless / mobile communication systems (such as 5G and New Radio (NR)) are unified networks / systems whose goal is to meet different and even conflicting performance dimensions and services. This diverse multi-dimensional requirement is driven by different services and applications. Typically, NR will evolve based on 3GPP LTE-Advanced, with the addition of potential new radio access. Further, it is expected that NR will evolve with the addition of potential new radio access technologies (RATs) to enrich mobile communications with improved, simple and seamless wireless connection solutions. NR can enable mobile communications that provide fast and rich content and services.

[0033] Technologies based on artificial intelligence / machine learning (AI / ML) can be used with 3GPP. AI / ML can be used in 5G evolution and 6G phases.

[0034] By considering BS-UE / UE-UE collaboration scenarios to support AI / ML operations, AI / ML can be used for RAN applications such as PHY, MAC, etc. AI / ML can facilitate collaborative-level interactive operations and data information flows to achieve AI / ML support for AI / ML-supported communication modes.

[0035] One or more embodiments are disclosed that facilitate RAN applications by using AI / ML operations for collaborative scenarios and / or communication modes.

[0036] The one or more embodiments include but are not limited to:

[0037] Includes machine learning (ML) parameter profiles for multiple domains to configure each parameter set for multiple communication modes under ML operation.

[0038] A UE ML capability profile including multiple domains to provide reference information about UE capabilities for performing ML operations.

[0039] Configuration of multiple communication modes and associated operational procedures.

[0040] UE cluster mapping method for selective codebook prioritization based on a single parameter.

[0041] UE cluster mapping method for combined weighted prioritization based on multiple parameters.

[0042] Functionality of Cluster Reference UE (CRU) and CRU selection process.

[0043] Communication mode triggering method using UE clustering.

[0044] A Gaussian mixture model (GMM) based quantization method is used for the codebook mapping process.

[0045] UE mobility based BS-BS cooperation method for combined training and split training.

[0046] UE mobility based triggering method for ML operations for model training.

[0047] Figure 1The architecture of a system 100 of a network according to some embodiments is shown. The system 100 is shown to include user equipment (UE) 101, 102, 103, and 104. The UEs 101 to 104 are shown as smart phones (e.g., handheld touch screen mobile computing devices that can connect to one or more cellular networks), but may also include any mobile or non-mobile computing devices, such as personal digital assistants (PDAs), pagers, laptop computers, desktop computers, wireless handheld devices, automotive equipment (e.g., vehicles), or any computing device that includes a wireless communication interface.

[0048] In some embodiments, any of UEs 101 to 104 may include an Internet of Things (IoT) UE, which may include a network access layer designed for low-power IoT applications (which utilize short-term UE connections). The IoT UE may utilize technologies such as machine-to-machine (M2M) or machine-type communication (MTC) to exchange data with an MTC server or device via a public land mobile network (PLMN), proximity-based services (ProSe) or device-to-device (D2D) communication, a sensor network, or an IoT network. The M2M or MTC data exchange may be a machine-initiated data exchange. The IoT network describes IoT UEs interconnected with short-term connections, which may include uniquely identifiable embedded computing devices (within the Internet infrastructure). The IoT UE may execute background applications (e.g., keep-alive messages, status updates, etc.) to facilitate connectivity to the IoT network.

[0049] The UEs 101 to 104 may be configured to connect (e.g., be communicatively coupled) to radio access networks (RANs) 111 and 112, which may be, for example, evolved universal mobile telecommunications system (UMTS) terrestrial radio access networks (E-UTRANs), NextGen RANs (NG RANs), or some other type of RANs. The UEs 101 to 104 are wirelessly connected to the BSs, and the air interface technology may be based on a cellular communication protocol, such as a global system for mobile communications (GSM) protocol, a code division multiple access (CDMA) network protocol, a push-to-talk (PTT) protocol, a cellular PTT (POC) protocol, a universal mobile telecommunications system (UMTS) protocol, a 3GPP long term evolution (LTE) protocol, a fifth generation (5G) protocol, a new radio (NR) protocol, etc.

[0050] In this embodiment, UE 101 to 104 can further directly exchange communication data via a sidelink interface, which includes one or more logical channels, including but not limited to a physical sidelink control channel (PSCCH), a physical sidelink shared channel (PSSCH), a physical sidelink discovery channel (PSDCH), and a physical sidelink broadcast channel (PSBCH).

[0051] An access node (AN) may be referred to as a base station (BS), NodeB, evolved NodeB (eNB), next generation NodeB (gNB), RAN node, etc., and may include a ground station (e.g., a terrestrial access point) or a satellite station providing coverage within a geographic area (e.g., a cell). A network device as referred to herein may include any of these APs, ANs, UEs, or any other network components.

[0052] In this embodiment, CN 121 provides the functionality to communicate with the UE, store its subscriptions and credentials, allow access to external networks and services, provide security, and manage network access and mobility.

[0053] The AN may include circuitry (eg, baseband circuitry), memory, a network interface (eg, an RF interface), one or more processors, and the like.

[0054] Figure 2 is a diagram illustrating example AI / ML module usage in a radio access network according to one or more embodiments.

[0055] It should be appreciated that BS-UE / UE-UE / BS-BS collaboration supports AI / ML operations in the RAN.

[0056] In typical previous BS-UE / UE-UE / BS-BS communication behaviors, support for AI / ML operations was not considered.

[0057] Depending on different AI / ML training / models, BS-UE / UE-UE / BS-BS communication modes need to be specified to achieve support.

[0058] Different configurations of BS-UE / UE-UE / BS-BS communication modes can enhance AI / ML operation performance. Examples of communication modes that can enhance this performance include network access mode, sidelink mode, broadcast / multicast mode, etc.

[0059] Figure 3 is a diagram illustrating federated learning operations using an AI / ML module in a radio access network according to one or more embodiments.

[0060] By collecting local model update feedback from each UE, the global model in the BS is distributed to each UE.

[0061] In this scenario, AI / ML training operations between the BS and multiple UEs have potential challenges, such as large signaling traffic and increased device power consumption.

[0062] Multiple communication modes supporting AI / ML operations include UE clustering scenarios and non-UE clustering scenarios.

[0063] UE clustering is based on AI / ML operations in UE clustering scenarios.

[0064] UE mobility is based on AI / ML operations in both UE clustering scenarios and non-UE clustering scenarios.

[0065] Figure 4 An upper table and a lower table are shown, which illustrate multiple communication modes according to one or more embodiments.

[0066] The above table 601 includes a mode index and defines an associated cooperative formation type of a BS / UE, such that an AI / ML operation is performed in a different manner based on the mode index.

[0067] To determine the communication mode(s) supporting AI / ML operations, a predefined ML parameter profile index configuration is used for mode selection. Additional details of the ML parameter profile are described below.

[0068] The following table 602 relates to the characteristics of the modes indicating different combinations in terms of signaling traffic load and device power consumption.

[0069] In case of AI / ML operation based on mode selection, improvements in signaling traffic load can be achieved.

[0070] The level or amount of improvement is based on the domain, such as communication domain, device domain, training domain, task domain, application domain, etc.

[0071] Figure 5 ML parameters for AI / ML operations according to one or more embodiments are presented.

[0072] Generate ML parameter profiles to provide multi-domain parameter sets so that appropriate communication mode selections can be made to support AI / ML operations in the RAN.

[0073] The parameter profile index configuration includes the following: communication domain, device domain, training domain, task domain, and application domain.

[0074] The information listed in each field of the ML parameter profile may be obtained from different sources located in application servers, core networks, radio access networks, UE devices, edge computing services, etc.

[0075] It is also possible to predefine multiple thresholds for parameters used for communication mode switching.

[0076] 701 shows an example of generating an ML parameter profile index for communication mode selection.

[0077] A UE ML capability profile may be generated to provide reference information about the UE's capabilities for performing (multiple) AI / ML operations. The profile may also be used for UE cluster formation and cluster reference assignment criteria for multi-communication mode selection.

[0078] When the RRC is established, the UE ML capability profile is provided to the node (BS, MEC, etc.) through the UE Capability Information message.

[0079] A UE measurement report message may also be sent to the BS to provide communication domain information of the UE ML capability profile.

[0080] Figure 6 A UE ML capability profile is presented in accordance with one or more embodiments.

[0081] 702 shows an example of generating (multiple) UE ML capability profiles.

[0082] Fig. 7A , Figure 7B , Figure 7C and Fig.7D is a diagram illustrating various multi-communication modes for supporting AI / ML operations according to one or more embodiments.

[0083] Fig. 7A CM-1 is shown.

[0084] Figure 7B CM-2 is shown.

[0085] Figure 7C CM-3 is shown.

[0086] Fig.7D CM-4 is shown.

[0087] In CM-1 mode, the BS communicates with a group of UEs for AI / ML operations.

[0088] The location of the AI / ML global model used for training depends on the specific implementation. For example, the global model can be located in BS, CN, MEC or cloud server, or UE, etc.

[0089] In this assumption, the AI / ML global model can be located in the BS. However, the global model can also be located elsewhere, such as in the edge computing network, or the upper network, or a central server, or any (multiple) dedicated UEs.

[0090] Each UE is capable of performing AI / ML local model training.

[0091] For AI / ML global model training, a subset of UEs are selected among the candidate UEs linked to the BS for (multiple) pre-configured training rounds based on the ML parameter profile index as the selection criteria.

[0092] The AI / ML global model is trained based on a collection of AI / ML local model training update feedback from the UE.

[0093] In CM-2, the BS communicates with the UE cluster to perform AI / ML operations.

[0094] The location of the AI / ML global model used for training depends on the specific implementation.

[0095] In this assumption, the AI / ML global model can be located in the BS. However, the global model can also be located in the edge computing network, or the upper network or the central server.

[0096] Each UE is capable of performing AI / ML local model training.

[0097] UE cluster(s) are formed for communicating with the BS, and each cluster is assigned a cluster reference UE (CRU) that represents the cascaded uplink transmission of AI / ML local model updates by collecting AI / ML local model updates from the clustered UE members.

[0098] The AI / ML global model is trained based on a collection of AI / ML local model training update feedback from the UE cluster.

[0099] It is also possible to selectively select available UE clusters from available UE clusters for AI / ML model training rounds based on AI / ML global model convergence performance or UE cluster status information and ML parameter configuration file updates. In other words, not all UE clusters need to participate in AI / ML model training rounds.

[0100] The selection criteria for CRUs are specified individually.

[0101] In CM-3, sidelink communication is possible between UE clusters for AI / ML operations.

[0102] The exact location of the AI / ML global model used for training depends on the specific implementation.

[0103] In this assumption, the AI / ML global model can be located in the BS. However, the global model can also be located in the edge computing network, or the upper network or the central server.

[0104] Each UE is capable of performing AI / ML local model training.

[0105] The AI / ML cluster model can also optionally be used as an alternative to the global model if desired.

[0106] Each cluster has its own cluster model for AI / ML training led by a cluster reference UE (CRU), and optionally shares cluster model updates or collected local model updates with the BS and / or neighboring clusters.

[0107] The AI / ML global model is trained based on a collection of AI / ML local model training update feedback from the UE cluster.

[0108] Optionally, cluster-level AI / ML model training can also be made to update each other if necessary.

[0109] Each cluster's CRU shares the following information with other clusters:

[0110] Cascaded local model updates generated in each cluster and / or

[0111] Global model updates forwarded from the BS to any cluster(s) but not received by them for their own local model training updates.

[0112] In CM-4, multiple BSs communicate with each other to perform AI / ML operations.

[0113] The exact location of the AI / ML global model used for training depends on the specific implementation.

[0114] In this assumption, the AI / ML global model can be located in the BS. However, the global model can also be located in the edge computing network, or the upper network or the central server.

[0115] Each UE is capable of performing AI / ML local model training.

[0116] UE clustering can also be combined into this communication mode.

[0117] The AI / ML cluster model can also optionally be used as an alternative to the global model if desired.

[0118] The BSs are connected to each other and to neighboring BSs through backhaul links, and AI / ML model information is exchanged to support AI / ML model training based on BS-BS combination.

[0119] The AI / ML model training performed in each BS can be shared with each other.

[0120] If necessary, the trained AI / ML model in each BS can also be used for transfer learning.

[0121] When neighboring BSs are unable to perform global model training (due to resources, energy, traffic load, or lack of trainable UEs), the leading BS may optionally include UE-local training model updates from other BSs in the neighboring BSs.

[0122] Figure 8 is a flow chart depicting a method of operating a communication mode utilizing AI / ML operations according to one or more embodiments.

[0123] At 1002, ML parameter profile status and UE ML capability profile updates are monitored.

[0124] At 1004 , an activation request for an AI / ML operation is detected.

[0125] Parameter sets for the ML parameter profile are prioritized based on the AI / ML operation 1006 .

[0126] At 1008, a communication mode is determined based on the prioritized set of parameters having predefined constraint thresholds.

[0127] At 1010, the determined communication mode is activated to initiate AI / ML operation support.

[0128] At 1012 , a check is performed for a communication mode switch. If yes, the method proceeds to 1006 .

[0129] Fig. 9 is a diagram illustrating a signaling flow for AI / ML operations utilizing a communication mode according to one or more embodiments. The operations are provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0130] The figure includes a plurality of UEs linked to BS1, BS1, BS2, and a plurality of second UEs linked to BS2.

[0131] The first UE and BS1 perform RRC setup. Additionally, the UE provides BS1 with UE measurement results / ML capability profile reports.

[0132] Similarly, the second UE and BS2 perform RRC setup. Additionally, the second UE provides BS2 with UE measurement results / ML capability profile reports.

[0133] BS1 and BS2 monitor the ML parameter profile status and UE ML capability profile update.

[0134] BS1 detects the activation request for AI / ML operation. Then, BS1 determines the communication mode (CM-1, CM-2, CM-3 or CM-4) based on the parameter priority ranking.

[0135] BS1 generates and provides an initial AI / ML operation configuration to the first BS1. Then, the first UE provides UE codebook related information as feedback.

[0136] Additionally, BS1 and BS2 share AI / ML operation configuration.

[0137] BS1 determines UE clustering using CRU selection, and BS2 activates BS-BS communication mode using a triggering event.

[0138] BS1 and the first UE develop a training setup based on UE clustering. BS1 shares AI / ML global model updates with BS2.

[0139] BS1 provides AI / ML global model updates to the first UE. In response, the first UE provides AI / ML local model updates.

[0140] BS2 shares the AI / ML global model update with the second UE.

[0141] In response, the second UE provides the AI / ML local model update to BS2.

[0142] BS2 provides the AI / ML local model update from the second UE to BS1.

[0143] BS1 updates the AI / ML global model.

[0144] BS1 completes AI / ML global model training after convergence.

[0145] The first UE and the second UE are released for the training connection.

[0146] Fig.10 is a flow chart illustrating a method for a CRU according to one or more embodiments. The operations are provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0147] The method includes the functionality of a Cluster Reference UE (CRU). The CRU represents the UE members of the cluster. The CRU cascades AI / ML local model updates to feed back to the BS or other clusters. The device type can vary, such as mobile electronic devices, vehicles, mobile edge computing, etc.

[0148] CRU executes AI / ML cluster model (optionally).

[0149] The AI / ML cluster model is trained based on the aggregation of AI / ML local model updates generated by the clustered UE members.

[0150] CRU selection is based on CRU equipment capabilities, and the CRU capability criteria is based on a specific implementation of the UE ML capability profile.

[0151] The priority of the UE and the service priority within the UE are also part of the CRU selection criteria.

[0152] At 1202, UE ML capability profile information is received.

[0153] At 1204, UE clustering is performed based on a clustering criterion.

[0154] At 1206, the initial UE is assigned as a CRU for each cluster based on the UE ML capability profile.

[0155] At 1208, additional CRU candidate UE(s) are determined when the initial CRU is to be handed over to another UE in its cluster.

[0156] At 1210, the AI / ML global model is distributed to cluster UE members.

[0157] At 1212, the AI / ML local model is updated from the cluster UE members to feed back into the AI / ML global model training.

[0158] Fig.11 is a flow chart illustrating a method for a CRU according to one or more embodiments. The operations are provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0159] The method describes setting a communication mode trigger using UE clustering. UE clustering is used to trigger (multiple) communication modes that enable (multiple) UE clusters supporting AI / ML operations.

[0160] The determination of the communication mode for AI / ML operation is based on the ML parameter profile update and the UE ML capability profile information.

[0161] Meanwhile, when CRU selection and UE clustering are available, then the UE cluster-based communication mode can be activated.

[0162] In this scenario, UE clustering is used as a triggering function to activate the associated communication mode.

[0163] At 1302, ML parameter profile update status and UE ML capability profile information are monitored.

[0164] At 1304, available UE clusters are identified based on the CRU selection.

[0165] At 1306, a check is performed to see if UE cluster(s) / CRUs are available.

[0166] If yes at 1306 , then at 1308 , a UE clustering-based communication mode supporting AI / ML operations is triggered.

[0167] At 1310, formation of UE clusters is activated.

[0168] Fig.12 is a diagram showing a signaling flow of a mode adopting UE cluster communication according to one or more embodiments. The operation is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0169] The figure is described with respect to a base station (BS) and one or more UE / CRUs. As shown, the flow may proceed in order from top to bottom, however, it should be understood that suitable variations in the order of execution are contemplated.

[0170] Perform and signal RRC setup.

[0171] UE measurement results / ML capability profile report is provided by UE / CRU.

[0172] BS initiates AI / ML operation configuration.

[0173] The BS provides the AI / ML initial operation configuration to the UE / CRU.

[0174] The UE / CRU provides codebook related information feedback to the BS.

[0175] The BS determines UE clusters and the CRU of each cluster.

[0176] The BS provides the sidelink resource configuration / cluster information to the UE / CRU.

[0177] Perform AI / ML model training (downlink (DL): global model, uplink (UL): local model).

[0178] The BS monitors UE mobility for mode switching and / or re-clustering.

[0179] The UE / CRU provides sidelink feedback to the BS.

[0180] The BS determines the mode switch and / or CRU change.

[0181] Fig.13is a graph depicting a set of Gaussian distributions according to one or more embodiments. The graph is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0182] It should be understood that UE clustering is basically based on the ML parameter profile index and the UE ML capability profile information.

[0183] For efficient AI / ML global model training operations, each UE cluster may have a reduced or minimized unbalanced data distribution of the local data sets of the cluster UE members.

[0184] For this purpose, a Gaussian Mixture Model (GMM) is used to compensate for the unbalanced data distribution of the local data sets of the cluster UE members.

[0185] Based on local data distribution information from each UE, an initial number of clusters is set, and then each UE is assigned to a cluster index, wherein the UEs have the same cluster index selected.

[0186] At the same time, by generating multi-codebook based quantization, two or more parameters can be used as decision criteria for UE clustering. In this case, UE local data distribution can be one of the parameters, and multiple parameters can be considered together before assigning a UE to any cluster index.

[0187] Fig.15 The Gaussian distribution shown in depicts the data distribution of the local data set of the cluster UE members. For this example:

[0188] Using GMM, we define a certain number of Gaussian distributions (GD i , where i=1, ..., C), and each of these distributions represents a cluster, where the clustered UE members have similar local data distributions based on the quantization method.

[0189] The set of Gaussian distributions used for quantization is denoted as GD i :{μ i , σ i}, as the {mean, variance} of the i-th group.

[0190] First, a “codeword” is defined as a quantized value representing the data distribution, as X i (the i-th codeword).

[0191] Then, the set of all codewords is defined as a "codebook".

[0192] A quantization technique for forming clusters is also described. For this technique, when clustering UEs with (multiple) parameter similarities, the number of UE clusters needs to be limited to be configurable. A limited amount of parameter data is determined to represent all available data distributions of candidate cluster UE members for mapping. By adopting any single parameter for the quantization codebook or multiple parameters for multiple codebooks, decision criteria for UE clustering can be formulated. The mapping criteria can be based on quantization error measurements. In this way, the signaling traffic overhead in UE cluster-based AI / ML operations (such as federated learning) can be reduced.

[0193] Fig.14 is a diagram depicting AI / ML operations based on single parameter UE clustering in accordance with one or more embodiments. The diagram is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0194] This figure shows selective codebook prioritization based on a single parameter for UE cluster mapping.

[0195] Fig.15 is a diagram depicting AI / ML operations for multiple parameter UE clustering in accordance with one or more embodiments. The diagram is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0196] This figure shows a combined weighted prioritization based on multiple parameters for UE cluster mapping.

[0197] Two techniques are described for UE cluster mapping: selective codebook prioritization based on a single parameter and combined weighted prioritization based on multiple parameters.

[0198] Single Parameter Based Selective Codebook Prioritization Starting from UE measurement data for a parameter set, a selected single parameter based codebook is used to assign UEs to each cluster with parameter prioritization.

[0199] Combined weighted prioritization based on multiple parameters Starting from UE measurement data of a parameter set, a selected multi-codebook based on multiple parameters is used to assign UEs to each cluster with parameter prioritization.

[0200] Fig.16 is a diagram depicting the operation of a BS and a UE for clustering using a codebook scheme according to one or more embodiments. The diagram is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0201] Perform and signal RRC setup.

[0202] The UE sends UE capability information and / or measurement results.

[0203] The BS configures the ML parameters and the UE ML capability profile.

[0204] The BS sends AI / ML operation configuration information to the UE.

[0205] The UE sends the codebook-based local data distribution to the BS.

[0206] The BS determines the clustering of UEs using CRU selection.

[0207] BS sends Cluster ID / CRU notification and RRC reset.

[0208] The BS monitors the cluster status / CRU status to cluster updates or changes.

[0209] Fig.17 is a flow chart describing a method of selecting a CRU for each cluster according to one or more embodiments. This figure is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0210] Codebook clustering starts at 1902, where the BS receives UE ML capability profile information and configuration of ML parameter profile index from the UE.

[0211] At 1904, a codebook having a set of codewords is generated.

[0212] At 1906, each UE is assigned to one of the predefined codewords indicating a cluster index based on the UE local data distribution.

[0213] At 1908, a CRU is selected for each cluster based on the selection criteria.

[0214] Fig.18 is a diagram depicting UE mobility using BS-BS combined training in accordance with one or more embodiments. The diagram is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0215] In AI / ML operations utilizing a radio access network, UEs involved in AI / ML local model training have mobility by moving between different BSs. AI / ML operation performance (such as model training) may be degraded due to convergence issues. UE mobility under AI / ML operations can be managed to mitigate or resolve this issue. In one example, a federated learning process is used.

[0216] In order to manage the degradation of AI / ML operation performance due to UE mobility, BS-to-BS combined training or split training can be used. To this end, the BSs are connected to each other with neighboring BSs via backhaul links. AI / ML model information is exchanged to support BS-BS combined / split training. The AI / ML model training updates performed in each BS can be shared with each other.

[0217] If necessary, the trained AI / ML model in each BS can also be used for transfer learning.

[0218] UE mobility can also trigger BS-BS based communication mode and can activate AI / ML operations across connected BSs by sharing the executed AI / ML model updates.

[0219] exist Fig. 20 Combined training is shown as an example in When performing AI / ML global model update, the BS collects AI / ML local model updates from UEs linked to (multiple) neighboring BSs.

[0220] Possible scenarios:

[0221] When UEs move between different BSs for connection, if BS-BS combined training is supported, AI / ML local model training updates from those UEs can continue to be provided to facilitate AI / ML global model updates.

[0222] Any single BS has a small number of UEs for AI / ML local model training, while with a sufficient number of UEs it may be eligible to perform BS-BS combined training.

[0223] Fig.19 is a diagram depicting UE mobility using BS-BS split training in accordance with one or more embodiments. The diagram is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0224] When each base station has their own connected UEs, two or more BSs share the same AI / ML global model for model training.

[0225] Possible scenario: When UEs move between different BSs for connectivity, if BS-BS split training is supported, AI / ML local model training updates from those UEs can continue to be provided to facilitate AI / ML global model updates.

[0226] Depending on the AI / ML capability status of the BS, the BS may optionally perform AI / ML model training in parallel with neighboring BSs, so that UE local model training with wider coverage can be obtained for AI / ML global model updates.

[0227] Fig. 20 is a flow chart illustrating a method of using BS-BS combined training and / or BS-BS split training according to one or more embodiments. The figure is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0228] Method 2200 depicts BS-BS combined training.

[0229] At 2202, an ML parameter profile index configuration and UE ML capability update are determined.

[0230] At 2204, neighboring BSs requesting BS-BS combined training for the AI / ML global model are identified.

[0231] At 2206, the initial AI / ML global model is distributed to UEs including other UEs linked to neighboring BSs.

[0232] Method 2250 depicts BS-BS segmentation training.

[0233] At 2252, an ML parameter profile index configuration with UE ML capability update is determined.

[0234] At 2254, it is determined whether a neighboring BS is needed to train the AI / ML global model.

[0235] At 2256, neighboring BSs capable of BS-BS segmentation training for the AI / ML global model are identified.

[0236] At 2258, the initial AI / ML global model is distributed to neighboring BSs for BS-BS segmentation training.

[0237] At 2260, AI / ML global model training is performed in parallel across different BSs.

[0238] At 2262, the AI / ML model update(s) are exchanged and iterated until target convergence is achieved.

[0239] Fig.21 is a flow chart illustrating a method for UE mobility-based triggering according to one or more embodiments. The figure is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0240] UE mobility is used to trigger the communication mode that enables BS-BS combination or split training for AI / ML operation. The determination of the communication mode for AI / ML operation is based on the ML parameter profile update and the UE ML capability profile information. At the same time, when any UE movement occurs during AI / ML operation (such as model training), then the communication mode based on UE mobility can be activated. In this scenario, UE mobility is used as a trigger function to activate the associated communication mode to support BS-BS combination / split training.

[0241] At 2302, ML parameter profile update status and UE ML capability profile information are monitored.

[0242] At 2304, any UEs in motion that are participating in AI / ML operations for model training are identified.

[0243] At 2306, the mobility of the identified UE(s) is checked.

[0244] If so, the BS-BS communication mode is triggered at 2308 .

[0245] At 2310, BS-BS combined training or split training is determined for activation.

[0246] Fig. 22 is a diagram illustrating global model sharing according to one or more embodiments. The diagram is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0247] Based on using UE mobility as a communication mode trigger for AI / ML operations for model training, the BS can share the AI / ML global model with neighboring BSs for training (as BS-BS split training) or collect AI / ML local model updates from UEs linked to neighboring BSs (as BS-BS combined training).

[0248] BS1 can share the global model with BS2.

[0249] BS2 may provide local model updates from UEs that have moved from BS1 to BS2.

[0250] Fig.23 is a diagram illustrating global model sharing according to one or more embodiments. The diagram is provided for illustrative purposes, and it should be understood that suitable variations are contemplated.

[0251] The figure depicts a UE connected to BS1, BS1 and BS2.

[0252] BS1 initiates AI / ML global model training.

[0253] BS1 sends AI / ML global model updates to UEs connected to BS1.

[0254] The UE responds with AI / ML local model update(s).

[0255] A UE among these UEs decides to move to BS2 based on the handover condition.

[0256] The UE sends a handover notification to BS1.

[0257] BS1 activates the BS-BS communication mode to perform combining / splitting training.

[0258] The handover process to BS2 begins.

[0259] The UE moves to BS2 and connects.

[0260] BS2 sends AI / ML global model updates to UE.

[0261] UE sends AI / ML local model updates to BS2.

[0262] BS2 updates the global model based on the AI / ML local model updates from the UE.

[0263] BS2 provides AI / ML local model updates to BS1.

[0264] BS1 updates the AI / ML global model using the AI / ML local model from BS2.

[0265] BS1 stops BS-BS global model training based on target convergence.

[0266] A system of one or more computers may be configured to perform specific operations or actions by installing software, firmware, hardware, or a combination thereof on the system that, in operation, causes the system to perform actions or causes the system to perform actions. One or more computer programs may be configured to perform specific operations or actions by including instructions that, when executed by a data processing device, cause the device, system, etc. to perform these actions.

[0267] A general aspect includes a system (100) having a base station (BS). The system also includes a radio frequency (RF) interface; and one or more processors configured to: monitor machine language (ML) parameter profile states (701) and ML capability profiles (702) of user equipment (UE) (101, 102) linked to the BS; detect activation requests (1006) for AI / ML operations; determine a communication mode based on parameter prioritization; share AI / ML operation (1006) configuration with a second base station (BS2) using the RF interface; determine UE clustering and determine a cluster reference UE (CRU); receive a UE local model from the CRU using the RF interface; receive a BS2 local model from the BS2 using the RF interface; update an AI / ML global model based on the UE local model and the BS2 local model; and release the UEs for training connections.

[0268] Implementations may include one or more of the following features. The system further includes a UE, which may include: a radio frequency (RF) interface, a memory, and one or more processors, the one or more processors being configured to: generate a capability profile; generate an ml parameter profile; generate a UE local model; and perform training using the UE local model and / or the ai / ml global model. The one or more processors of the BS are configured to generate configurations of multiple communication modes and associated operation flows. The one or more processors of the BS are configured to provide a communication mode triggering method using UE clustering. The one or more processors of the BS are configured to generate a Gaussian mixture model (GMM)-based quantization method for a codebook mapping process. The one or more processors of the BS are configured to: receive a machine language (ML) parameter profile, which may include multiple parameter sets for multiple domains; configure the multiple parameter sets for multiple communication modes under ML operation. The one or more processors of the BS are configured to generate a UE mobility-based triggering method for ml operations for model training. The capability profile may include multiple domains to provide reference information about UE capabilities for performing ML operations. Implementations of the described technology may include hardware, methods or processes, or computer software on a computer-accessible medium.

[0269] A general aspect includes an apparatus for a base station (BS) (111). The apparatus also includes a radio frequency (RF) interface; and one or more processors configured to: receive a machine language (ML) parameter configuration file, the parameter configuration file may include multiple parameter sets for multiple domains; and configure the multiple parameter sets for multiple communication modes under ML operation.

[0270] Implementations may include one or more of the following features. The one or more processors are configured to receive a UE ML capability profile, which may include multiple domains for providing reference information about UE capabilities for performing ML operations. The one or more processors are configured to perform a BS-BS collaboration method based on UE mobility for combined training and split training. The one or more processors are configured to generate a triggering method based on UE mobility for ML operations for model training. The one or more processors are configured to generate configurations of multiple communication modes and associated operating procedures. The one or more processors are configured to provide a communication mode triggering method using UE clustering. The one or more processors are configured to generate a quantization method based on a Gaussian mixture model (GMM) for a codebook mapping process. Implementations of the described technology may include hardware, methods or processes, or computer software on a computer-accessible medium.

[0271] A general aspect includes an apparatus for a base station (BS) (111). The apparatus also includes a radio frequency (RF) interface; and one or more processors configured to: receive UE (101) measurement results and capability profile reports of multiple user equipment (UE) via the RF interface; determine one or more UE clusters of the multiple UEs; determine a cluster reference UE (CRU) for each of the one or more clusters; monitor UE mobility of the multiple UEs; and determine a changed CRU for the one or more clusters based on the monitored UE mobility.

[0272] Implementations may include one or more of the following features. The one or more processors are configured to identify available UE clusters based on the determined CRU. The one or more processors are configured to trigger a UE clustering-based communication mode that supports AI / ML operations. Implementations of the described technology may include hardware, methods or processes, or computer software on a computer-accessible medium.

[0273] One general aspect includes one or more computer-readable media having instructions. The medium also includes monitoring ML parameter profiles and UE ML capability profile information. The medium also includes identifying one or more UEs that are in motion and participate in AI / ML model training. The medium also includes determining to perform model training with a second base station (BS2).

[0274] These embodiments may include one or more of the following features. The one or more computer-readable media have instructions that, when executed, cause the BS to further perform the following operations: detecting an activation request for AI / ML operation; determining a communication mode based on parameter prioritization; and sharing an AI / ML operation configuration with a second base station (BS2). Embodiments of the described technology may include hardware, methods or processes, or computer software on a computer-accessible medium.

[0275] As used herein, the term "circuitry" may refer to, components of, or include an application specific integrated circuit (ASIC), an electronic circuit, a processor (shared, dedicated, or group), and / or memory (shared, dedicated, or group) that executes one or more software or firmware programs, a combinational logic circuit, and / or other suitable hardware components that provide the described functionality. In some embodiments, the circuitry may be implemented in one or more software or firmware modules, or functions associated with the circuitry may be implemented by one or more software or firmware modules. In some embodiments, the circuitry may include logic that is at least partially operable in hardware.

[0276] As used in this specification, the term "processor" may refer to substantially any computing processing unit or device, including but not limited to a single-core processor; a single processor with software multithreaded execution capability; a multi-core processor; a multi-core processor with software multithreaded execution capability; a multi-core processor with hardware multithreading technology; a parallel platform; and a parallel platform with distributed shared memory. In addition, a processor may refer to an integrated circuit, an application-specific integrated circuit, a digital signal processor, a field programmable gate array, a programmable logic controller, a complex programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions and / or processes described herein. The processor may utilize nanoscale architectures, such as, but not limited to, transistors, switches, and gates based on molecules and quantum dots, in order to optimize space usage or enhance the performance of mobile devices. The processor may also be implemented as a combination of computing processing units.

[0277] In this specification, terms such as "storage", "data storage", "data storage device", "database", and any other information storage components substantially related to the operation and functionality of the components and / or processes refer to "memory components", or entities embodied in "memory" or components that include memory. It should be noted that the memory components described herein may be volatile memory or non-volatile memory, or may include both volatile memory and non-volatile memory.

[0278] By way of illustration and not limitation, for example, non-volatile memory may be included in memory, non-volatile memory (see below), disk storage (see below), and memory storage (see below). Further, non-volatile memory may be included in read-only memory, programmable read-only memory, electrically programmable read-only memory, electrically erasable programmable read-only memory, or flash memory. Volatile memory may include random access memory, which acts as an external cache memory.

[0279] It should be understood that the various aspects described herein can be implemented by hardware, software, firmware or any combination thereof. When implemented with software, the function can be transmitted or stored on a computer-readable medium as one or more instructions or codes. Computer-readable media include both computer storage media and communication media, and communication media include any media that facilitates the transfer of computer programs from one place to another. Storage media or computer-readable storage devices can be any available media that can be accessed by general or special computers. By way of example and not limitation, such computer-readable media can include RAM, ROM, EEPROM, CD-ROM or other optical disk storage devices, disk storage devices or other magnetic storage devices, or other tangible and / or non-transient media, which can be used to carry or store desired information or executable instructions. In addition, any connection is properly referred to as a computer-readable medium. For example, if a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL) or wireless technologies such as infrared, radio and microwave are used to transmit software from a website, server or other remote source, then coaxial cable, fiber optic cable, twisted pair, DSL or wireless technologies such as infrared, radio and microwave are included in the definition of medium. As used herein, disk and disc include compact disc (CD), laser disc, optical disc, digital disc (DVD), floppy disk and Blu-ray disc, wherein disks usually reproduce data magnetically, while discs reproduce data optically using lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0280] The various illustrative logics, logic blocks, modules, and circuits described in conjunction with the aspects disclosed herein may be implemented or performed using a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in an alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, for example, a combination of a DSP and a microprocessor, a plurality of microprocessors, a combination of one or more microprocessors and a DSP core, or any other such configuration. In addition, at least one processor may include one or more modules operable to perform one or more steps and / or actions described herein.

[0281] For software implementation, the technology described herein can be implemented with a module (e.g., program, function, etc.) that performs the functions described herein. The software code can be stored in a memory unit and executed by a processor. The memory unit can be implemented within the processor or outside the processor, in which case the memory unit can be communicatively coupled to the processor by various means known in the art. In addition, at least one processor may include one or more modules that are operable to perform the functions described herein.

[0282] The technology described herein can be used for various wireless communication systems, such as CDMA, TDMA, FDMA, OFDMA, SC-FDMA and other systems. The terms "system" and "network" are usually used interchangeably. A CDMA system can implement radio technologies such as Universal Terrestrial Radio Access (UTRA), CDMA1800, etc. UTRA includes other variants of Wideband-CDMA (W-CDMA) and CDMA. Further, CDMA1800 covers IS-1800, IS-95 and IS-856 standards. A TDMA system can implement radio technologies such as Global System for Mobile Communications (GSM). An OFDMA system can implement radio technologies such as Evolved UTRA (E-UTRA), Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.18, Flash-OFDM, etc. UTRA and E-UTRA are part of Universal Mobile Telecommunications System (UMTS). 3GPP Long Term Evolution (LTE) is a version of UMTS that uses E-UTRA, which employs OFDMA on the downlink and SC-FDMA on the uplink. UTRA, E-UTRA, UMTS, LTE, and GSM are described in documents from an organization named "3rd Generation Partnership Project (3GPP)". Additionally, CDMA1800 and UMB are described in documents from an organization named "3rd Generation Partnership Project 2" (3GPP2). Further, such wireless communication systems may additionally include peer-to-peer (e.g., mobile-to-mobile) ad hoc network systems, which typically use unpaired unlicensed spectrum, 802.xx wireless LAN, BLUETOOTH, and any other short-range or long-range wireless communication technology.

[0283] Single carrier frequency division multiple access (SC-FDMA) utilizing single carrier modulation and frequency domain equalization is a technique that may be used with the disclosed aspects. SC-FDMA has similar performance and substantially similar overall complexity to OFDMA systems. SC-FDMA signals have lower peak-to-average power ratio (PAPR) due to their inherent single carrier structure. SC-FDMA may be used in uplink communications, where lower PAPR may benefit mobile terminals in terms of transmit power efficiency.

[0284] In addition, the various aspects or features described herein can be implemented as methods, devices or products using standard programming and / or engineering techniques. The term "product" used herein is intended to include a computer program that can be accessed from any computer-readable device, carrier or medium. For example, computer-readable media can include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical disks (e.g., compact disks (CDs), digital versatile disks (DVDs), etc.), smart cards, and flash memory devices (e.g., EPROMs, cards, sticks, key drives, etc.). In addition, the various storage media described herein can represent one or more devices and / or other machine-readable media for storing information. The term "machine-readable medium" can include, but are not limited to, wireless channels and various other media capable of storing, containing and / or carrying (multiple) instructions and / or data. In addition, a computer program product can include a computer-readable medium having one or more instructions or codes that can be operated to cause a computer to perform the functions described herein.

[0285] Communication media embodies computer readable instructions, data structures, program modules, or other structured or unstructured data as a data signal, such as a modulated data signal, such as a carrier wave or other transport mechanism, and includes any information delivery or transmission media. The term "modulated data signal" or multiple signals refers to a signal that has one or more characteristics of the signal set or changed in a manner that encodes information in one or more signals. By way of example, and not limitation, communication media includes wired media, such as a wired network or direct-wired connection, and wireless media, such as acoustic, RF, infrared and other wireless media.

[0286] In addition, the actions of the methods or algorithms described in conjunction with the aspects disclosed herein may be directly embodied in hardware, software modules executed by a processor, or a combination thereof. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An example storage medium may be coupled to a processor so that the processor can read information from a storage medium and write the information to the storage medium. In an alternative, the storage medium may be integrated into the processor. Further, in some aspects, the processor and the storage medium may reside in an ASIC. In addition, the ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components. In addition, in some aspects, the steps and / or actions of the method or algorithm may reside in a machine-readable medium and / or a computer-readable medium that may be incorporated into a computer program product as one or any combination or set of codes and / or instructions.

[0287] The above description of the illustrated embodiments of the subject disclosure, including what is described in the Abstract, is not intended to be exhaustive or to limit the disclosed embodiments to the precise forms disclosed. Although specific embodiments and examples are described herein for illustrative purposes, those skilled in the relevant art may recognize that various modifications are possible within the scope of such embodiments and examples.

[0288] In this regard, while the disclosed subject matter has been described in conjunction with various embodiments and corresponding figures, it should be understood that other similar embodiments may be used or modifications and additions may be made to the described embodiments, where applicable, to perform the same, similar, alternative, or alternative functions of the disclosed subject matter without departing from the disclosed subject matter. Therefore, the disclosed subject matter should not be limited to any single embodiment described herein, but rather should be construed in breadth and scope in accordance with the claims appended hereto.

[0289] In particular, with respect to the various functions performed by the above-described components (assemblies, devices, circuits, systems, etc.), unless otherwise specified, the terms used to describe such components (including references to "means") are intended to correspond to any component or structure (e.g., functionally equivalent) that performs the specified function of the described component, even if not structurally equivalent to the disclosed structure that performs the functions in the exemplary embodiments shown in the present disclosure herein. In addition, while a particular feature may have been disclosed for only one of several embodiments, such feature may be combined with one or more other features of other embodiments, which may be desirable and advantageous for any given or specific application.

Claims

1. A system (100) having a base station (BS) (111, 112), comprising a circuit system having the following items: Radio Frequency (RF) interface; and One or more processors configured to: Monitoring a machine language (ML) parameter profile state (701) and a ML capability profile (702) of a user equipment (UE) (101, 102) linked to the BS; Detect activation request for AI / ML operation (1006); Determining a communication mode based on parameter prioritization; sharing the AI / ML operation (1006) configuration with a second base station (BS2) using the RF interface; Determine UE clustering and determine a cluster reference UE (CRU); receiving a UE local model from the CRU using the RF interface; receiving a BS2 local model from the BS2 using the RF interface; Updating the AI / ML global model based on the UE local model and the BS2 local model; and These UEs are released for training connections.

2. The system of claim 1, further comprising a UE, the UE comprising: Radio frequency (RF) interface; Memory; as well as One or more processors configured to: Generate capability profiles; Generate ML parameter configuration file; generating a local model of the UE; and Training is performed using the UE local model and / or the AI / ML global model.

3. The system of claim 1, wherein the capability profile comprises a plurality of fields for providing reference information on UE capabilities for performing ML operations.

4. The system of any one of claims 1 to 3, wherein the one or more processors of the BS are configured to generate configurations of multiple communication modes and associated operation procedures.

5. The system of any one of claims 1 to 4, wherein the one or more processors of the BS are configured to provide a communication mode triggering method using UE clustering.

6. The system of any one of claims 1 to 5, wherein the one or more processors of the BS are configured to generate a Gaussian mixture model (GMM) based quantization method for a codebook mapping process.

7. The system according to any one of claims 1 to 6, wherein the one or more processors of the BS are configured to: receiving a machine language (ML) parameter configuration file, the parameter configuration file comprising a plurality of parameter sets for a plurality of domains; The multiple parameter sets are configured for multiple communication modes under ML operation.

8. The system of any one of claims 1 to 6, wherein the one or more processors of the BS are configured to generate a UE mobility-based triggering method for ML operations for model training.

9. An apparatus for a base station (BS) (111), comprising a circuit system having the following items: Radio Frequency (RF) interface; and One or more processors configured to: receiving a machine language (ML) parameter configuration file, the parameter configuration file comprising a plurality of parameter sets for a plurality of domains; The multiple parameter sets are configured for multiple communication modes under ML operation.

10. The apparatus of claim 9, the one or more processors being configured to receive a UE ML capability profile, the capability profile comprising a plurality of fields for providing reference information on UE capabilities for performing ML operations.

11. The apparatus of claim 9, the one or more processors being configured to generate a configuration of multiple communication modes and associated operational flows.

12. The apparatus of claim 9, the one or more processors configured to provide a communication mode triggering method using UE clustering.

13. The apparatus of claim 9, the one or more processors configured to generate a Gaussian mixture model (GMM) based quantization method for a codebook mapping process.

14. The apparatus of any one of claims 9 to 13, the one or more processors being configured to perform a BS-BS cooperation method based on UE mobility for combined training and split training.

15. The apparatus of any one of claims 9 to 13, wherein the one or more processors are configured to generate a UE mobility-based triggering method for ML operations for model training.

16. An apparatus for a base station (BS) (111), comprising a circuit system having the following items: Radio Frequency (RF) interface; and One or more processors configured to: receiving user equipment (UE) (101) measurement results and capability profile reports of a plurality of user equipments (UEs) via the RF interface; determining one or more UE clusters of the plurality of UEs; determining a cluster reference UE (CRU) for each of the one or more clusters; monitoring UE mobility of the plurality of UEs; as well as An altered CRU is determined for the one or more clusters based on the monitored UE mobility.

17. The apparatus of claim 16, the one or more processors configured to identify available UE clusters based on the determined CRU.

18. The apparatus of claim 16, the one or more processors configured to trigger a UE clustering-based communication mode for AI / ML operations.

19. One or more computer readable media having instructions that, when executed, cause a base station (BS) to: Monitor ML parameter profile and UE ML capability profile information; Identify one or more UEs that are in motion and participate in AI / ML model training; It is determined to perform model training with the second base station (BS2).

20. The one or more computer-readable media of claim 19 having instructions that, when executed, cause the BS to further perform the following operations: Detect activation requests for AI / ML operations; determining a communication mode based on the parameter prioritization; and The AI / ML operation configuration is shared with the second base station (BS2).

Images