Ue context configuration for persistence during AI / ML operation

Abstract

There is provided an apparatus comprising: one or more processors and a memory storing instructions that, when executed by the one or more processors, cause the apparatus to: submit, by a source network entity of a RAN, a request to a target network entity, the request comprising at least a User Equipment Machine Learning (UEML) group index, where the UE ML group index indicates that the UE is part of a UE ML group; a response message is received by the source network entity from the target network entity based at least on the indicated UE ML group index.

Description

Technical Field

[0001] The present application relates to a method and system for improving user equipment (UE) context management related to optimizing handover procedures, and more specifically to maintaining artificial intelligence (AI) machine learning (ML) context in a radio access network (RAN) between at least a source network entity and a target network entity (such as two next generation NodeBs, gNBs). Background Art

[0002] Any discussion of the background art throughout the specification should not be considered as an admission that this art is widely known or forms part of the common general knowledge in the field.

[0003] The advent of the fifth generation technology standard (5G) for broadband cellular networks has prompted the study of new use cases and potential service requirements for 5G systems to also support the use of AI / ML to enhance services.

[0004] Therefore, a new SI RP-201620 "NR and EN-DC Data Collection Enhancement Study" was launched in RAN3#110e, with the goal of specifically studying the high-level principles for launching AI in RAN and the functional framework including the inputs and outputs required by AI functions and ML algorithms, so as to achieve high-precision outputs and possible feedback optimization in 5G systems. Specifically, the SI aims to identify the data required for AI functions in the data generated in the input and output, as well as the impact of standardization on nodes or network interfaces in the existing architecture to transmit the above input / output data through them.

[0005] Based on this, various solutions are proposed in 3GPP TR 37.817 to initially implement AI / ML implementations in RAN. In these solutions, different network interfaces are usually introduced, including at least "input data for training" (training data), "input data for reasoning" (reasoning data), output information after making ML decisions, and feedback required to capture the effect of the corresponding ML model, which leads to at least a functional framework concept for AI / ML use in RAN.

[0006] However, despite ongoing efforts to improve the accuracy and stability of ML models currently implemented in RAN, a problem remains: due to the dynamic and distributed architecture of the RAN (hence the need for the source gNB to request and obtain training or measurement data from other (target) gNB nodes), at least the basic information enabling entity-specific AI / ML model outputs or overall inter-node-based RAN actions may be lost, resulting in a significant reduction in the efficiency of the corresponding RAN processes.

[0007] For example, one method for receiving neighbor node information is to utilize the resource status procedure. However, the resource status information does not include any further specifications about, for example, the individual UEs from which the requested data is obtained, which results in that in the current RAN architecture signaling, there is no way to collect UE (or UE group) specific information from neighbor nodes, let alone store this information for use after the UE context is released.

[0008] Therefore, it is necessary to propose a new method and apparatus for a RAN-based feedback system, especially a RAN-based AI / ML model system, which solves some or all of the above problems in an efficient, flexible and reliable manner. Summary of the invention

[0009] According to some aspects, the subject matter of the independent claims is provided. Some further aspects are defined in the dependent claims.

[0010] According to a first aspect of the present application, at least a first device is provided, comprising:

[0011] one or more processors; and

[0012] a memory storing instructions that, when executed by the one or more processors, cause the first device to:

[0013] Submitting, by a source network entity of a radio access network RAN, a request to a target network entity, the request comprising at least a user equipment machine learning UE ML group index, wherein the UE ML group index indicates that the UE is part of a UE ML group; and / or

[0014] A response message is received by the source network entity from the target network entity based at least on the indicated UE ML group index.

[0015] Here, in some examples, the corresponding target network entity and the source network entity may be connected via RAN. In addition, at least the source network entity may be a first NG-RAN node, and the target network entity may be a second NG-RAN node connected to the source network entity or operation management and maintenance (OAM), which results in the following effects: By the above instructions, an inter-node signaling operation in the RAN may be generated, which involves exchanging UE-specific information, thereby exchanging potential UE ML contexts.

[0016] In addition, in some examples, the request may include an ID and configuration information for identifying a UE ML group, wherein the ID is configured for at least one of: indicating requested data to be collected by the target network entity, or indicating data collected by the target network entity required for training and / or inference of a machine learning ML model.

[0017] Therefore, through the above-mentioned ID, the measurement data required for using the AI ​​ML model in the RAN can be initially specified, resulting in faster and resource-saving acquisition of the corresponding stored measurement data from the requested UE or target network entity. As a result, in some examples, the target network entity can be configured to query and receive only ID-specific measurement data from each predefined UE, or the target network entity can be configured to query and receive all measurement data from each predefined UE, and then filter out only the ID-specific measurement data. Similarly, the source network entity can be configured to request and receive the corresponding collected measurement data from the target network entity based on the corresponding selected ID, thereby selectively and efficiently aggregating the required information by calculating performance information corresponding to a certain ID that may be useful, for example, for implementing AIML model training at the source network entity.

[0018] In a preferred example of the present application, a UE ML group is set before the handover process and is preferably only available / accessible between network nodes during inter-node signaling. A UE ML group may characterize UEs with common reasoning actions for providing feedback information, even after the handover is completed, where the reasoning actions may be only a subset. For example, due to UE history or following similar movement paths or trajectories, radio capabilities, activated services or slices, radio conditions, UEs affected by network actions (e.g., SON or AI / ML algorithm output or some other network decision), etc., a UE ML group may indicate a group of UEs that receive similar treatment in the RAN.

[0019] Here, in some examples, the measurement data collected from the UE may correspond to the output and parameters of non-limiting UE measurement values, such as UE measurement values ​​related to RSRP, RSRQ, SINR of the serving cell or neighboring cell of the corresponding UE. Alternatively, at least when making an AI ML decision, the measurement values ​​(and therefore the available measurement output data) may also include data generated during reasoning from the target or source network entity or from the UE. In addition, the measurement values ​​and measurement data may also include UE-level performance information (e.g., bit rate, packet loss rate, latency, energy efficiency, etc.) on each UE affected by at least one reasoning action, wherein the reasoning action may be at least the triggering of a UE's handover due to an AI ML optimization decision (such as a mobility optimization decision), or shutting down a UE connected to a capacity cell for energy saving or load balancing purposes, to name just a few examples. Further optionally, the measurement data may also include cell-level performance information after taking a reasoning action, wherein the reasoning action may be based on multiple system key performance indicators (KPIs), such as throughput, latency, radio link failure (RLF), counters, etc. )

[0020] In addition, in some examples, the configuration information may also include instructions on how the requested data should be collected, wherein the requested data may include at least one of the following: a requested measurement value, a requested counter, or a requested prediction value. The request data preferably includes feedback information introduced for different use cases, such as: for energy saving use cases: resource status of neighboring NG-RAN nodes, energy efficiency, UE performance affected by energy saving actions (e.g., Ue of handover), including bit rate, packet loss rate, delay; system KPI (e.g., throughput, delay, current and neighboring RLF of NG-RAN nodes); for load balancing use cases: UE performance information from the target NG-RAN (for those Ue handed over from the source NG-RAN node), resource status information updates from the target NG-RAN, system KPI (e.g., throughput, delay, current and neighboring RLF); for mobility optimization use cases: QoS parameters, such as throughput, packet delay, etc. of UE of handover, resource status information updates from the target NG-RAN, and / or performance information from the target NG-RAN. In addition, in some examples, the feedback information may include timing information, such as information averaged over at least a period of time. Here, the feedback information may for example include data on a throughput or delay metric of the RAN, which is a metric averaged over a period of time after the handover is completed at the target network entity, so that the UE context no longer exists. In addition, the feedback information may also include information about the time period, such as a time window over which the average is calculated or a start time at which the corresponding measurement value needs to be calculated. Additionally or optionally, in addition to the start time, a start event may also be included, such as when the first UE in the UE ML group is handed over. All or part of this time-related information may be included in the instruction specifying the timing information of the requested data.

[0021] In some examples, the first device is further caused to: train and / or execute the ML model based at least on the received response message of the target network entity. Therefore, in the corresponding examples of the present application, both AI ML model training and / or model reasoning can be performed based on the implementation of UE ML group index specific data. In addition, since the output of the AI ​​ML model still depends on the fed data determined by the implemented UE ML group index, subsequent RAN signaling processes (such as feedback reporting or iterative feedback loops for AI ML model adjustment) can similarly rely on UE ML group index selection.

[0022] As a result, in some examples, in use cases involving UE handover, the response message may also be a feedback report provided after performing ML model inference.

[0023] In some examples, the device is caused to perform a UE handover procedure according to an optimization action based at least on information of the received response message. The optimization action may be the result of optimization using an optimization algorithm (such as AI / ML model reasoning), and the covered use cases may include energy saving, load balancing, and / or mobility optimization. More details on the use cases can be found in TR37.817, which is incorporated herein by reference.

[0024] Additionally, in some examples, the response message may include a UE ML group index and feedback information, where the feedback information may also include information of UEs affected by the ML reasoning.

[0025] In some examples, the UE ML group index may also include information about predefined network actions. For example, the UE ML group index may include information about the network operation for which the UE ML group index was created, which results in additional parameters that can be used for inter-node signaling in the RAN. For example, the UE ML group index may indicate the purpose of the AI ​​ML model of the collected measurement data (e.g., energy saving, load balancing, mobility enhancement, etc.). In other examples, the UE ML group index may also include information about the predefined criteria for which the UE ML group index was created, wherein the information may be read by at least one of the following source network entity or target network entity.

[0026] Thus, in some examples, the time period may include at least a time and / or event after a handoff or handoff procedure is completed at the target network entity.

[0027] To this end, at least one of the source network entity or the target network entity may be configured to locally store information about at least one UE ML group in a UE ML group context. The group context may be, for example, a UE ML group index, so that each of the corresponding target network entity and / or the source network entity may provide UE ML group context information. In addition, at least one of the source network entity or the target network entity may be configured to indicate that after a UE context release is sent from the target network entity to the source network entity, the UE ML group context should still exist (at least for a predefined or variable time).

[0028] As a result, in some examples, the first apparatus further causes: indicating, by the source network entity, that the UE ML group index is to remain unchanged after at least receiving a response message from the target network entity.

[0029] In some examples, the feedback information may be recorded in an ML report that accumulates requested data from neighboring network entities.

[0030] In some examples, the UE ML group index may assign UEs to a UE ML group based on at least a predefined criterion, where the UE ML group may correspond to a group of UEs that receive similar treatment in the network. In some examples, the predefined criterion may be defined by at least one of: all UEs considered by the ML model, UEs with the same or similar UE history or following similar mobile paths or trajectories in the RAN, radio capabilities, activated services or slices, radio conditions, all UEs whose RSRP values ​​are below a threshold or all UEs whose RSRP values ​​are above a threshold, UEs affected by network actions based on ML decisions or outputs, UEs associated with a network entity, a specific beam index or cell ID connected to a distributed unit (DU), or UEs belonging to a defined area.

[0031] In some examples, the request to the target network entity requests information for each UE ML group. In addition, the response message from the target network may provide information for each UE ML group.

[0032] In some examples, the first device can be further configured to send a configuration message to at least one UE, which includes configuration information for initiating predefined UE measurements; and receive a measurement report message from at least one UE, which includes the output of the predefined UE measurements.

[0033] Alternatively, according to the second aspect of the present application, a second device may also be provided, including:

[0034] one or more processors; and

[0035] at least one memory storing instructions that, when executed by one or more processors, cause the apparatus to:

[0036] The source network entity transmits a handover request message to the target network entity of the RAN, wherein the handover request message includes a UE ML group index, feedback configuration information, and an instruction, wherein the instruction indicates that feedback information should be available after the handover is completed for the UE and / or feedback information should be available after a context release message is sent from the target network entity to the source network entity;

[0037] A response message is received by the source network entity from the target network entity, which is based at least on the indicated UE ML group index and includes the feedback information.

[0038] In addition, alternatively or additionally, the device may be caused to transmit a handover request message from a source network entity to a target network entity, the message including a user equipment machine learning UE ML group index and an instruction indicating that the feedback information should be available after the handover is completed for the UE or the feedback information should be available after a context release message is sent from the target network entity to the source network entity. After calculating (average) performance information (e.g., throughput, delay, QoS, etc.), a context release message may be sent from the target network entity to the source network entity. When the target network entity sends feedback information to the source network entity, the context release may be indicated to the source by the target network entity. When the source network entity receives feedback information corresponding to a specific AI ML context from the target network entity, it may interpret the indication to release the context. The source node may also independently trigger the context release when the source node no longer needs to maintain a specific AI ML context corresponding to the UE ML group with the target network entity or when an internal timer expires. The context release message may be sent at a later time after the handover is performed and after the UE context of the handover operation no longer exists. In addition, the source network entity may be configured to store information of the message required for handover at least until a release message (preferably, for example, a context release message sent from the target network entity) is received at the source network entity. In addition, the apparatus may also enable the source network entity to receive a response message from the target network entity, which is at least based on the indicated UE ML group index and includes the feedback information; and perform the UE handover process according to the optimization action based at least on the feedback information of the received response message.

[0039] In a further development, a device is proposed, which, as a target network entity, receives a handover required message, which includes a user equipment machine learning UE ML group index and an instruction, the instruction indicating that feedback information should be available after the handover is completed for the UE or after a context release message is sent from the target network entity to the source network entity. The target network entity can preferably be further configured to transmit a response message based at least on the indicated UE ML group index and including the feedback information to the source network entity; and based at least on the feedback information of the received response message, perform the UE handover process according to an AI / ML optimization action (or alternatively a non-AI / ML optimization action, and therefore an optimization based on a non-AI optimization algorithm). The feedback information (in the case of an AI / ML optimization application) can also be used as input data for training (such as training data) and / or input data for reasoning (such as reasoning data).

[0040] In addition, according to the third aspect of the present application, a third device may also be provided, including:

[0041] one or more processors; and

[0042] at least one memory storing instructions that, when executed by the one or more processors, cause the third device to:

[0043] receiving, at a target network entity of the RAN, a request including at least a UE ML group index, wherein the UE ML group index may indicate that the UE is part of a UE ML group;

[0044] collecting and locally storing feedback information about UEs in an existing UEML group of the target network entity based at least on the UEML group index; and

[0045] Transmitting a response message to the source network entity based on the indicated UE ML group index. In some examples, the request may be an ML context initiation request, which includes configuration information indicating requested data to be collected by the target network entity or a feedback report request for information of UEs affected by ML reasoning.

[0046] In some examples, in a use case involving UE handover, the response message may be a feedback report provided after performing ML model inference that includes information of UEs affected by the ML inference.

[0047] In addition, according to the fourth aspect of the present application, a fourth device may also be provided, including:

[0048] one or more processors; and

[0049] at least one memory storing instructions that, when executed by the one or more processors, cause the fourth device to:

[0050] In a use case involving UE handover, a feedback report provided after executing the ML model action is transmitted by the target network entity of the RAN,

[0051] The feedback report is transmitted upon receiving a request from the source network entity, or when the feedback report is available at the target network entity or at a predetermined time after the handover.

[0052] According to another aspect of the present application, a method of a first device may also be provided, wherein the first device includes at least one or more processors and a memory storing instructions, and the method includes:

[0053] submitting, by a source network entity of the first device of the RAN, a request to a target network entity, the request comprising at least a UE ML group index, wherein the UE ML group index indicates that the UE is part of a UE ML group; and,

[0054] A response message is received by the source network entity from the target network entity based at least on the indicated UE ML group index.

[0055] According to another aspect of the present application, a method of a second device may also be provided, where the second device includes at least one or more processors and a memory storing instructions, and the method includes:

[0056] transmitting, by a source network entity of the second device, a handover request message to a target network entity of the RAN, the message including a UE ML group index, feedback configuration information and an instruction, the instruction indicating that the feedback information should be available after handover of the UE is completed or after a context release message is sent from the target network entity to the source network entity; and

[0057] A response message is received by the source network entity from the target network entity, which is based at least on the indicated UE ML group index and includes the feedback information.

[0058] According to another aspect of the present application, a method of a third device may also be provided, where the third device includes at least one or more processors and a memory storing instructions, and the method includes:

[0059] receiving, at a network entity of a third device of the RAN, a request including at least a UE ML group index, wherein the UE ML group index may indicate that the UE is part of a UE ML group;

[0060] collecting and locally storing feedback information about an existing UEML group of the UE of the target network entity based at least on the UEML group index; and

[0061] Transmitting a response message to the source network entity based on the indicated UE ML group index.

[0062] According to another aspect of the present application, a method of a fourth device may also be provided, where the fourth device includes at least one or more processors and a memory storing instructions, and the method includes:

[0063] In a use case including UE handover, a feedback report provided after performing the ML model action is transmitted by a target network entity of said fourth means of the RAN,

[0064] The feedback report is transmitted upon receiving a request from the source network entity, or when the feedback report is available at the target network entity or at a predetermined time after the handover.

[0065] In some examples of at least one of the above methods of the first device to the fourth device, the feedback information may include information averaged over at least a period of time. Specifically, the period of time may include at least a time after the handover is completed at the target network entity. In a further advantageous development, the period of time may be defined in the ML context initiation request message and / or may be part of the measurement configuration.

[0066] In some examples of at least one of the above methods of the first to fourth devices, the feedback information used can be recorded in an ML report that accumulates request data from neighboring network entities.

[0067] In some examples of at least one of the above methods of the first device to the fourth device, the UE ML group index can assign the UE to the UE ML group based on at least a predefined criterion, wherein the UE ML group can correspond to a UE group receiving similar treatment in the network.

[0068] In some examples of at least one of the above methods of the first device to the fourth device, the predefined criteria can be defined by at least one of the following: all UEs considered by the ML model, UEs with the same or similar UE history or following similar mobile paths or trajectories in the RAN, radio capabilities, activated services or slices, radio conditions, all UEs whose RSRP values ​​are lower than a threshold or all UEs whose RSRP values ​​are higher than a threshold, UEs affected by network actions based on ML decisions or outputs, UEs associated with a network entity, a beam index or cell ID connected to a distributed unit (DU), or UEs belonging to a defined area.

[0069] In some examples of at least one of the above methods of the first device to the fourth device, the group index may include information about the predefined criteria for creating the UE ML group index, wherein the information can be read by at least one of the following source network entity or target network entity.

[0070] In some examples of at least one of the above methods of the first device to the fourth device, at least the method of the first device or the second device may further include: storing, by a source network entity of the first device or the second device, a UE ML group index at a predetermined time, wherein the predetermined time is at least a time after handover is completed at a target network entity. In addition, preferably, the UE ML group and / or index is only valid for feedback collection purposes, and after receiving the feedback, and if no other feedback is to be processed, it will be invalid or can be deleted.

[0071] In some examples of at least one of the above methods of the first device to the fourth device, at least the method of the first device or the second device may further include: indicating, by the source network entity of the first device or the second device, that the UE ML group index will be maintained after at least receiving a response message from the target network entity. In addition, preferably, the UE ML group and / or the UE ML group index is only effective for feedback collection purposes, and after receiving the feedback, if there is no other feedback to be processed, it will be invalid or can be deleted.

[0072] In some examples of at least one of the above methods of the first device to the fourth device, the method of at least the first device or the second device may also include: sending a configuration message to at least one UE, which includes configuration information for initiating predefined UE measurements; and receiving a measurement report message from at least one UE, which includes the output of the predefined UE measurements.

[0073] In some examples of at least one of the above methods of the first device to the fourth device, the request received by the target network entity of at least the third device or the fourth device may be an ML context initiation request, which includes configuration information indicating the target network entity to collect requested data or a feedback report request of information of the UE affected by the ML reasoning. In addition, in a further advantageous development, the configuration information of the ML context initiation request may also include timing information about when the measurement value collection at the target starts, when it stops, and / or which measurement values ​​should be collected.

[0074] In some examples of at least one of the above methods of the first device to the fourth device, in a use case including UE handover, the response message transmitted by the target network entity to the source network entity may be a feedback report including information of the UE affected by the ML reasoning and provided after performing the ML model reasoning.

[0075] In addition, according to a fifth aspect of the present application, a network system may also be provided, including:

[0076] at least one of the first device or the second device; and

[0077] at least one of a third device or a fourth device;

[0078] wherein at least one of the first device or the second device and at least one of the third device or the fourth device are connected via a RAN; and

[0079] The network system is configured to enable at least one of the first device or the second device and at least one of the third device or the fourth device to perform any of the above method steps.

[0080] In addition, according to the third aspect of the present application, a third device may also be provided, including:

[0081] one or more processors; and

[0082] at least one memory storing instructions that, when executed by the one or more processors, cause the third device to:

[0083] receiving, at a target network entity of the RAN, a request including at least a UE ML group index, wherein the UE ML group index may indicate that the UE is part of a UE ML group;

[0084] collecting and locally storing feedback information about UEs in an existing UEML group of the target network entity based at least on the UEML group index; and

[0085] A response message is transmitted to the source network entity based on the indicated UE ML group index.

[0086] Although UE ML group, UE ML group index and UE ML context are described, which may be features of an AI / ML framework, in further development, if a non-AI / ML algorithm is used, they may be replaced by UE group (or UE optimization group), UE group index (or UE optimization group index) and UE context (or UE optimization context).

[0087] In addition, according to some other example embodiments, for example, a computer program product for a wireless communication device is provided, the wireless communication device comprising at least one processor, which includes a software code portion for executing the corresponding steps disclosed in the present application when the product is run on the device. The computer program product may include a computer-readable medium having the software code portion stored thereon. In addition, the computer program product can be directly loaded into the internal memory of the computer and / or transmitted via a network by means of at least one of an upload, download, and push process.

[0088] Although some example embodiments will be described herein with particular reference to the above-mentioned applications, it should be understood that the present teachings are not limited to such fields of use and may be applied in a wider range of environments.

[0089] It is worth noting that it should be understood that the method according to the present application relates to a method of operating a device according to the above-mentioned example embodiments and variations thereof, and corresponding statements made for the device are also applicable to the corresponding method, and vice versa, so similar descriptions may be omitted for the sake of brevity. In addition, even if not explicitly disclosed, the above aspects may be combined in a variety of ways. Those skilled in the art will understand that the combination of these aspects and features / steps is possible unless it creates a contradiction that is explicitly excluded.

[0090] Implementation of the disclosed apparatus may include, but is not limited to, the use of one or more processors, one or more application specific integrated circuits (ASICs), and / or one or more field programmable gate arrays (FPGAs). Implementation of the apparatus may also include the use of other conventional and / or custom hardware, such as a software programmable processor, such as a graphics processing unit (GPU) processor.

[0091] It should be understood that any of the above modifications may be applied individually or in combination to the corresponding aspects to which they refer, unless other options are explicitly stated to be excluded.

[0092] Therefore, according to the content of the present application, the context loss problem can be solved by storing at least information related to the UE (especially at least necessary information for enabling entity-specific AI / ML model output or overall node-based RAN actions) for a duration beyond the initial identifiable duration of the UE in the RAN. In addition, beyond the duration that the UE can be identified in the RAN (especially in the source node), the information can be obtained from the neighboring nodes, so that the corresponding feedback and AI / ML processes can be efficiently initiated. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Further details, features, objects and advantages will be apparent from the following detailed description of preferred embodiments of the present application in conjunction with the accompanying drawings, in which:

[0094] Figure 1 shows the context activation initiation process between the source network entity and the target network entity;

[0095] Figure 2 shows a message exchange for initiating a feedback message between a source network entity and a target network entity;

[0096] Figure 3A Message exchanges to achieve UE context persistence in a general feedback-based action between a source network entity and a target network entity are shown;

[0097] Figure 3B shows the message exchange to achieve UE context persistence in the AI ​​ML process between a first NG-RAN node and a second NG-RAN node;

[0098] Figure 4 The message exchange to achieve UE context persistence in the AI ​​ML process between the first NG-RAN node and OAM is shown. DETAILED DESCRIPTION

[0099] Below, based on the communication network architecture of the 3GPP standard for communication networks (such as 5G / NR), different exemplary embodiments will be described using examples of communication networks to which the exemplary embodiments can be applied, but the embodiments are not limited to such architectures. It is obvious to those skilled in the art that these embodiments can also be applied to other types of communication networks, where the mobile communication principle is combined with D2D (device to device) or V2X (vehicle to vehicle) configurations, such as SL (side link)), such as Wi-Fi, Worldwide Interoperability for Microwave Access (WiMAX), Personal Communications Service (PCS), Wideband Code Division Multiple Access (WCDMA), systems using ultra-wideband (UWB) technology, mobile ad hoc networks (MANET), wired access, etc. In addition, without loss of generality, the description of some embodiments relates to mobile communication networks, but the principles of this application can be extended and applied to any other type of communication network, such as a wired communication network.

[0100] The following examples and embodiments should be understood as illustrative embodiments only. Although the specification may refer to "one", "an" or "some" examples or embodiments in multiple places, this does not necessarily mean that each such reference relates to the same example or embodiment, or that the feature applies only to a single example or embodiment. Individual features of different embodiments may also be combined to provide other embodiments. In addition, terms such as "including" and "comprising" should be understood as not limiting the described embodiments to consist only of the features already mentioned; these examples and embodiments may also include features, structures, units, modules, etc. that are not specifically mentioned.

[0101] The basic system architecture of a (telecommunication) communication network including a mobile communication system to which examples of some embodiments apply may include the architecture of one or more communication networks, including a radio access network subsystem and a core network. Such an architecture may include one or more communication network control elements or functions, access network elements, radio access network (RAN) elements, access service network gateways or base transceiver stations, such as base stations (BS), access points (AP), NodeBs (NBs), eNBs or gNBs, distributed units (DUs) or centralized / central units (CUs), which control the corresponding coverage areas or cells, and one or more communication stations (such as communication elements or functions, such as user equipment or terminal equipment, such as user equipment (UE)) or another device with similar functions, such as a modem chipset, chip, module, etc., which may also be part of a station, element, function or application capable of communication, such as a UE, an element or function that can be used for a machine-to-machine communication architecture, or as a separate element attached to such an element, function or application capable of communication, etc., capable of communicating via one or more communication beams through one or more channels to send several types of data in multiple access domains. In addition, core network elements or network functions may also be included, such as gateway network elements / functions, mobility management entities, mobile switching centers, servers, databases, etc.

[0102] The following description may provide further details of alternatives, modifications and variations:

[0103] -gNB includes, for example, a node that provides NR user plane and control plane protocol terminations to the UE, and is connected to the 5GC via the NG interface, for example, according to Section 3.2 of 3GPP TS 38.300 V16.6.0 (2021-06) incorporated by reference. In addition, the following description and proposed features of the present application are not limited to application to the indicated framework, but may also be applicable to other generations such as Long Term Evolution (LTE) technology, as well as eNB or other technologies such as LTE-Advanced (LTE-A) technology.

[0104] User equipment (UE) may include a wireless or mobile device, a device with a radio interface to interact with a RAN (Radio Access Network), a smartphone, an in-vehicle device, an IoT device, an M2M device, etc. Such a UE or device may include: at least one processor; and at least one memory including computer program code; wherein the at least one memory and the computer program code are configured to, together with the at least one processor, cause the device to perform at least certain operations, such as an RRC connection with the RAN. The UE is configured to, for example, generate a message (e.g., including a cell ID) to be transmitted to the RAN via radio (e.g., to arrive at a serving cell and communicate with it). The UE may generate and transmit and receive an RRC message including one or more RRC PDUs (packet data units).

[0105] - In addition, a handover may be defined as a connection handover of a UE from a predetermined source network entity (preferably a source NG-RAN node) to a target network entity (preferably a target NR-RAN node). Here, a handover may be performed regardless of whether there is a control plane connection between the source network entity and the target network entity or not.

[0106] The UE may (e.g., according to 3GPP TS 38.331 V16.5.0 (2021-06) Sections 42.1 and 4.4, incorporated by reference) have different states.

[0107] For example, when the RRC connection is established, the UE is in the RRC_CONNECTED state or the RRC_INACTIVE state.

[0108] In the RRC_CONNECTED state, the UE may:

[0109] oStore AS context;

[0110] o Unicast data transmission to / from UE;

[0111] o monitor a control channel associated with a shared data channel to determine whether data is scheduled for the data channel;

[0112] oProvide channel quality and feedback information;

[0113] oPerform neighbor cell measurements and measurement reporting.

[0114] The RRC protocol includes the following main functions:

[0115] o RRC connection control;

[0116] oMeasurement configuration and reporting;

[0117] o Establishment / modification / release of measurement configurations (e.g. intra-frequency, inter-frequency and inter-RAT measurements);

[0118] oSetting and releasing of measuring gaps;

[0119] oMeasurement report.

[0120] The general functionality and interconnection of the elements and functions (also depending on the actual network type) are well known to those skilled in the art and are described in the corresponding specifications, so for the sake of brevity, a detailed description thereof may be omitted here. However, it is worth noting that in addition to those described in detail below, several additional network elements and signaling links may be employed to communicate to or from elements, functions or applications, such as communication endpoints, communication network control elements (such as servers, gateways, radio network controllers) and other elements of the same or other communication networks.

[0121] The communication network architecture considered in the example embodiments may also communicate with other networks, such as a public switched telephone network or the Internet. The communication network may also support the use of cloud services for virtual network elements or their functions, where it should be noted that the virtual network portion of the telecommunications network may also be provided by non-cloud resources, such as an internal network, etc. It should be understood that the network elements and / or corresponding functions of the access system, core network, etc. may be implemented using any node, host, server, access node or entity, etc. suitable for such purpose. In general, the network functions may be implemented as dedicated hardware, as software instances running on dedicated hardware, or as virtualized functions instantiated on an appropriate platform, such as a cloud infrastructure.

[0122] In addition, network elements, such as communication elements (such as UE), terminal equipment, control elements or functions, such as access network elements (such as base stations / BS), gNBs, radio network controllers, core network control elements or functions, such as gateway elements, or other network elements or functions as described herein, and any other elements, functions or applications can be implemented by software (e.g., by a computer program product for a computer) and / or by hardware. In order to perform the corresponding processing, the corresponding device, node, function or network element used may include several devices, modules, units, components, etc. (not shown) required for control, processing and / or communication / signaling functions. Such devices, modules, units and components may include, for example: one or more processors or processor units, which include one or more processing parts for executing instructions and / or programs and / or for processing data; storage or memory units or devices for storing instructions, programs and / or data, used as a work area for processors or processing parts, etc. (e.g., ROM, RAM, EEPROM, etc.); input or interface devices (e.g., floppy disks, CD-ROMs, EEPROMs, etc.) for inputting data and instructions through software; user interfaces (e.g., screens, keyboards, etc.) for providing monitoring and operation possibilities to users; other interfaces or devices (e.g., wired and wireless interface devices, radio interface devices including, for example, antenna units, etc., devices for forming radio communication parts, etc.) for establishing links and / or connections under the control of processor units or parts; wherein the corresponding devices forming the interfaces (such as radio communication parts) may also be located at remote sites (e.g., radio heads or radio stations, etc.). It is worth noting that in this specification, a processing part should not be regarded only as representing a physical part of one or more processors, but can also be regarded as a logical division of the mentioned processing tasks performed by one or more processors. It should be understood that according to some examples, a so-called "fluid" or flexible network concept may be adopted, in which the operations and functions of a network element, a network function or another entity of the network may be performed in a flexible manner in different entities or functions (e.g., in nodes, hosts or servers). In other words, the "division of labor" between the network elements, functions or entities involved may vary from case to case.

[0123] Furthermore, when it comes to AI / ML models in the RAN, the ML model may be executed and / or trained at at least one RAN network entity (such as a NG-RAN node side or an OAM side). The aforementioned network entity may have one or more trained or to-be-trained models available, and the models are preferably configured to solve a predetermined problem, preferably an optimization problem. Furthermore, a given network entity may also have internally implemented non-ML algorithms (e.g., already present in the network entity). Thus, the corresponding network is able to indicate to the network entity which model should be used at any given time and when to activate the model.

[0124] On the other hand, the network entity may also indicate to other network entities or to the network in general whether it is an ML capable network entity. Furthermore, even if the UE has indicated to the network or other network entities that it is ML capable, it is possible that the network entity may not be able to perform ML over time, for example, if the network entity detects that, for example, the network connection has changed or the power of the network entity has dropped below a certain threshold. Therefore, the network entity is able to dynamically indicate its current ML capability within the network, and the network may also dynamically select other network entities to implement the ML model on.

[0125] Therefore, since the RAN is a distributed architecture, a given network entity on which an ML model is implemented may need to request and obtain necessary information, such as measurement and / or training data, from other network entities. Here, such an inter-node process may be generated, for example, by a resource status process between the model owning network entity and a second network entity.

[0126] On the other hand, since specific UE context is usually lost after UE handover, current ML model implementations in RAN (but also non-ML model implementations) may suffer from severe efficiency degradation due to the situation that subsequent feedback or subsequent processes (i.e., implementation of model outputs) may not be redirected to the UE they originally responded to. For example, one method for receiving neighbor node information is to utilize the resource status process. However, resource status information involves node-based reporting and does not include any further specifications, such as from which UE to obtain the requested data, which results in that in the current RAN architecture signaling, there is no way to collect UE (or UE group) specific information from neighbor nodes, let alone store this information for use after UE context release. As a result, according to current standards, the target gNB is not aware of the need to collect specific data for performance feedback during the handover process when receiving the handover request. The cause information element (IE) in the aforementioned handover request can indicate the cause value of the handover (e.g., reducing serving cell load, resource optimization, UE power saving, etc.), but it does not provide a mechanism for the target node to identify specific data collection requirements to perform corresponding actions, such as performance feedback. Note that performance feedback (also referred to as feedback for brevity) can be used to monitor the performance of AI / ML models when available.

[0127] Therefore, in order to avoid the above mentioned problems, it is proposed that UE performance measurements should be additionally signaled from the target gNB after the handover is completed. However, since the UE context is no longer present in the source gNB soon after the handover is performed, the above solution may only allow the provision of performance information (of one or more UEs) until the corresponding UE context release time, which may prohibit the collection of required feedback information (e.g., about the average throughput or delay metrics over a predefined (usually (much) larger than the UE context lifetime) period after the handover is completed at the target gNB).

[0128] Therefore, according to some example embodiments, before requesting feedback related to the corresponding UE of a RAN action (handover), a new UE ML context may be initiated at least between multiple network entities of the corresponding RAN, preferably at least between NG-RAN nodes or OAM. Here, the feedback may correspond to measurements, counters or information for evaluating the performance of a given RAN action.

[0129] In some example embodiments, for this, a UE ML group index may be generated to identify a new UEML context and allow the gNB to track the UE beyond the initial context release point. Here, a given group index may respectively correspond to and identify a predetermined group of UEs (so-called UEML groups) defined by characteristic criteria, such as characteristic mobility paths or trajectories, radio capabilities, activated services, radio conditions or impact on predefined network actions, resulting in that a UE ML context may be initiated again even after the initial context (i.e., after a handover procedure) is lost by assigning the group index to at least one or more network entities within a given RAN.

[0130] Therefore, in the exemplary embodiment, the process step of new context initiation can be implemented in the corresponding RAN process (e.g., AI ML process) by allocating and transmitting the UEML group index in and between network entities of the corresponding RAN (i.e., between the source network entity and the target network entity), wherein the new context initiation can be an asynchronous message that can be initiated and terminated at each stage of the process and can be used to associate measurement values ​​and other information with the UE ML context.

[0131] Thus, in some example embodiments, a new context initiation may be sent:

[0132] - before requesting at least one of feedback, measurements, counters, information gain related to the UE for a RAN baseline action or an action generated as a result of AI / ML model inference (e.g. handover, energy saving decision, load balancing, etc.); or

[0133] - Before a network entity trains a model or implements corresponding measurement data into a predetermined application, to allow the aforementioned network entity to receive specific information / measurements for training / implementation purposes.

[0134] In addition, in an additional embodiment of the present application, a given network entity may further configure the target network entity to indicate which information (measurements, counters, etc.) it needs to be informed of at least during the initiation of a new context (preferably, it is the node that starts the ML context initiation, which configures which measurements it wants to receive from neighbors). Exemplarily, such information may be an average throughput or delay measurement of a predetermined UE in the UEML group. In addition, if the measurement involves an average value, the network entity may also configure the start time at which the measurement needs to be started at the target network entity and the end time at which the measurement needs to be stopped at the target network entity. Alternatively, the network entity may configure an average window with a start time and an average value for calculating the average value for the target network entity. Optionally, the start time of notifying when the measurement needs to be performed may also be indicated by a start event (e.g., when the first UE in the UEML group is handed over). Optionally, the end time of notifying when the measurement needs to be stopped may also be indicated by an end event at the node performing the measurement (e.g., if one or more UEs participating in the UEML group are subsequently handed over to another node).

[0135] In an additional embodiment, in order to collect and coordinate the obtained feedback, in the case of utilizing the AI ​​ML model, for example as an AI ML report, a given network entity may also be configured to send corresponding feedback according to an additional RAN action report. In order to additionally ensure a persistent UE ML context, the RAN action report may preferably be tagged with a UE ML group index, so that the corresponding requesting network entity thereby knows which UEs the measurements in the report refer to. In a further embodiment, the UE ML group index may also include information about the corresponding network action, such as the AIML algorithm that created the UE ML group index, so that the corresponding use initiated by the new context can be efficiently traced back.

[0136] therefore, Figure 1A first exemplary embodiment of a new context initiation process between a source network entity (source) and (one or more) target network entities (targets) is shown. It is noteworthy that multiple target network entities may also be used. As described above, the corresponding network entity may be at least an NG-RAN node or an OAM of the corresponding RAN, but the network entity is not limited to these elements. Therefore, the source network entity and the target network entity may equally be part of an upper-level device in the RAN, wherein the source network entity may be part of a first device and the target network entity may be part of a second device. Similarly, at least one of the first or second device or the target or source network may include a device for monitoring, a device for supervision, a device for indicating and / or a device for requesting required information or data, so that the corresponding device or network entity may be configured to independently perform and execute the working process imposed by the RAN. Here, the device for monitoring, the device for supervision, the device for indicating or the device for requesting may be a monitoring device, a supervisory device, an indicating device or a requesting device, respectively. The device for monitoring, the device for supervision, the device for indicating and the device for requesting may be a monitor, an indicator and a requester, respectively. Optionally, the means for monitoring, the means for supervising, the means for instructing or the means for requesting may also be a monitoring processor, a supervisory processor, an instructing processor or a requesting processor, respectively.

[0137] Therefore, in an exemplary embodiment, a communication is performed between a source network entity and a target network entity (or multiple target network entities). Figure 1 During the new context initiation process shown, first, a context initiation request (more preferably, an ML context initiation request) may be sent from a source network entity to one or more target network entities. The context initiation request may include initiation information, such as at least one UE ML group index being sought and / or a predefined measurement configuration, which indicates to the requested target network entity which information or measurement data stored by it is required by the source network entity, and how such information or data should be collected, for example, by indicating timing information (e.g., a start time / event, an end time / event, or an averaging window). In addition, the target network entity receiving the context initiation request may respond by sending back a context initiation response (more preferably, an ML context initiation response), which is a message that at least confirms (or rejects) a previous request of the source network entity.

[0138] Alternatively, instead of or in addition to the above embodiments, the context initiation request also need not be processed independently. Instead, in other exemplary embodiments, the context initiation request may also be piggybacked in other RAN actions, such as a handover preparation procedure, by which the source network entity may request the target network entity to collect additional data for the UE participating in the RAN action in a UE-associated manner. To this end, the source network entity may also include an indication that, in a given RAN action, after the initial UE context release is sent from the target network entity to the source network entity (receiving a new UE ML context via UE ML group index collection), the UE ML context will still persist, resulting in that for a particular UE, more data may subsequently be sent between the source network entity and the target network entity without risking the loss of a given UE ML context. Similarly, the source network entity may also indicate to the target network entity a data collection configuration that should be enabled at the target network entity after the UE completes a given handover.

[0139] In some examples, the functionality and processing steps of the new context initiation process described above may be applicable to the Xn interface (in the case of an Xn complete handover (HO)). In other examples, this may also be applicable to the NG interface, for example, in the case of an NG handover, by handover signaling via the core network (CN).

[0140] Furthermore, the actual measured values ​​of a given RAN action may also be sent in the feedback message (or, in the case of an AI ML model implementation, in the ML report), preferably as the measurement data is progressively collected. Figure 2 An exemplary embodiment of such a process is shown.

[0141] To this end, Figure 1Compared to the process shown in the figure, the corresponding source network entity can initially request a given feedback report by sending a feedback report request to a predetermined target network entity. The feedback report request may include the UE ML group information (e.g., index) for which feedback is requested as a parameter. As a result, the target network entity can respond to the aforementioned feedback report request and send the requested measurement data back to the source network entity in a feedback report. Therefore, since the above process steps can be performed by additionally considering the predetermined UE ML group index (thereby ensuring that there is a UE ML context in the whole process), an efficient UE-specific data request can be generated. Operations on UE ML groups can be collectively created, modified, and deleted, thereby considering all UEs of the UE ML group in an efficient manner. Modifications to UE ML groups may include adding or deleting measurements required for the group of UEs. Therefore, modifications to the UE ML group may affect all UEs of the group. Using the UE ML group index, neighboring nodes can identify and receive measurement information about the corresponding UE even after the context expires.

[0142] Furthermore, in a further example, feedback may be sent, in particular upon request or when a RAN action report (e.g., ML report) is available at a network entity (such as an NG-RAN node). Optionally, the feedback message may also be sent at a predefined time, in particular at an expected time window or time point.

[0143] Furthermore, it is again noted that the above new context initiation procedure may be handled for any given action (preferably optimization action) in the RAN, which may need to rely on feedback or other network procedures that need to outlast the initial UE ML context release (e.g., after a handoff). Thus, while the described embodiments may be implemented, for example, for AI ML optimization cases, such as, for example, using an AI ML model in the RAN for mobility optimization, energy saving, or load balancing, the corresponding new context initiation procedure may not be limited to such RAN actions, but may be used for any given RAN action that may require a prolonged UE context.

[0144] therefore, Figure 3A An exemplary message exchange for implementing UE context (UE ML context or UE optimization context) persistence in a general feedback-based action between a source network entity and a target network entity based on the first embodiment of the present application is shown.

[0145] Wherein, in step S100, a given target network entity may first perform a previous, preferably optional work process, such as additional feedback implementation, AI ML model processing, or any other process in the RAN, which may generate the required input, such as resource status and / or utilization prediction / estimation. In a preferred embodiment, such a work process may also be to initiate the next iteration loop generated by returning the feedback / output data of the corresponding RAN action generated in the last iteration step to the corresponding network entity. Therefore, the process described subsequently may also be understood as an iterative process or optimization loop, which may be processed a predefined number of times or until a given threshold amount is reached.

[0146] In step S101, the source network entity may then configure measurement information on the UE side and send a configuration message to the corresponding UE, which includes configuration information for instructing the UE to collect requested and UE-specific information data (e.g., performance data). At this time, the number of connected UEs is not limited. Similarly, the corresponding source network entity may be configured to send a configuration message not only to one UE but also to multiple UEs.

[0147] In step S102, the corresponding UE may then collect the indicated information data through UE-related measurements. Here, the UE measurements may, for example, relate to RSRP, RSRQ, SINR or serving cell and / or neighboring cell.

[0148] Subsequently, in step S103, the UE may send a measurement report message including the requested measurement value / information data back to the source network entity.

[0149] In response, in step S104, the source network entity may initiate a UE context for data that may be used for a given operation in the RAN by sending a context initiation request to the target network entity, as previously described in Figure 1 Here, the context initiation request may include an ID identifying the UE group identified by the predetermined UE group index so that the measurement data used can be subsequently associated with the UE used for measurement extraction. In addition, the ID can be used to indicate the measurement data to be implemented in the RAN or at the corresponding source network entity for the corresponding action (e.g., integrated application process). These measurement values ​​may then need to be collected by the target network entity to which the context initiation request is sent. The measurement configuration can also be given, for example, by indicating timing information (e.g., start time / event, end time / event or averaging window) to identify the requested measurement values ​​and how they should be collected.

[0150] Furthermore, in step S105 , the corresponding target network entity may then respond with a message confirming the UE context initiation for the requested measurement values.

[0151] In step S106, the source network entity may then obtain input data to be implemented for the corresponding RAN action from the target network entity, wherein the corresponding input data may include at least the required input information from the target network entity. Wherein, if the corresponding target network entity also performs a previous initiation process of the RAN action (e.g., model training of a given AI ML model), the input data may also include output data, such as an inference result from the target network entity. In addition, the input data received from the target network entity may be equivalently associated with the initiating UE context ID.

[0152] In optional step S106Z, an optimization process is performed. The optimization is to optimize the handover process, wherein the optimization preferably uses an optimization algorithm. The optimization algorithm can be modified based on feedback information received from the target network entity (for example, using a context initiation response and / or a feedback report). Since the feedback report information may include information about UEs affected by the optimization, UEs assigned to a UE group or a UE ML group, the algorithm can be further improved by taking into account the feedback information for modification. The feedback information may be available to the source network entity at least some time after the actual handover has ended, so that at least due to the assigned UE group or UE ML group and the related index, the identification of the UE is still possible at least during this time. In another example, the optimization algorithm may be an AI / ML algorithm or model.

[0153] Then, in step S107, based on the received input data and / or measurement results, at least one of the corresponding UEs, the source network entity or the target network entity may perform a corresponding RAN action including an integrated handover procedure (handover to a new cell) to handover the UE from at least the source network entity to the target entity. In one option, the RAN action may be parameterized by a UE context ID for this purpose, resulting in, for example, that the target network entity can know the information required for a given UE based on the measurement values ​​indicated in the UE context. Therefore, the target entity knows in this way that it should calculate (average) performance information (e.g., throughput, delay, QoS, etc.) including the UEs participating in the handover. If the UE context ID is not included in the RAN action, the UEs participating in the handover cannot be used to calculate (average) performance information. As an example, the RAN action may include at least a signaling procedure between the source and target network entities, which provides a handover request (step S107a) and a handover confirmation (Ack) signal (S107b) between the source and target network entities. Likewise, in order to avoid configuration errors between the network entity and the corresponding UE, a radio resource control (RRC) reconfiguration may be performed by providing a corresponding reconfiguration protocol from at least one of the source network entity or the target network entity to the corresponding UE (step S107c). As a result, it may be confirmed that the RRC reconfiguration is completed after a given handover (step S107d). Finally, upon completion of the RAN action and thus the corresponding handover, the source network entity may request feedback information related to the UE context ID from the target network resource in step S108, such as Figure 2 shown.

[0154] Therefore, in step S109, the target network entity may send the requested feedback information associated with the UE context to the source network entity, so that even after the handover is completed, the source network entity can still identify the corresponding UE of the predefined UE ML group. Therefore, using the given information, the corresponding source network entity can, for example, implement the output of the RAN action (e.g., the output of the AI ​​ML model) or provide corresponding feedback to an additional network entity or UE for additional optimization process. In addition, as described above, the generated feedback information can, for example, be restored to the RAN, specifically to the corresponding source or target network entity, so as to be Figure 3A The context persistence process shown restarts at step S100, generating an iterative optimization loop. Preferably, the feedback information is associated with the UE ML context ID.

[0155] also, Figure 3B Shows Figure 3BAdditional embodiments of the context persistence procedure, where the corresponding RAN actions can be explicitly structured as AI ML optimization procedures, such as mobility optimization procedures, energy saving or load balancing procedures, to name a few examples integrated in the NG-RAN node.

[0156] Therefore, with Figure 3A Compared to the rather general process shown in , the corresponding source network entity and the target network entity can be defined as predetermined NG-RAN nodes, respectively NG-RAN nodes 1 and 2, between which the corresponding handover can be performed.

[0157] Furthermore, as a result of the AI ​​ML model process, prior to the actual optimization action (shown here in step S211) including the corresponding handoff process, the model training ( Figure 3B , step S207) and model reasoning ( Figure 3B In the embodiment, step S210) can be added to the AI ​​ML optimization process to efficiently generate corresponding optimization solutions in the RAN.

[0158] As a result, the input data previously obtained from the target network entity (NG-RAN node 2) in step S206 at the source network entity (now NG-RAN node 1) can be used to train the corresponding AI ML model, wherein the input data used for training can also include the required input information from NG-RAN node 2. Optionally, if NG-RAN node 2 is to execute the AI ​​ML model, the input data used for training can also include the corresponding inference results from NG-RAN node 2. In addition, the input data received from NG-RAN node 2 can be associated with the originating UE context ID.

[0159] Then, in Figure 3B In step S207, model training may be initiated. Here, the required measurements may be utilized in the corresponding AI ML model for predefined optimization reasons, such as the aforementioned mobility optimization, energy saving, or load balancing optimization.

[0160] In addition, Figure 3B In step S208, after the model training is completed, the corresponding NG-RAN node 1 can also obtain the measurement report as additional reasoning data for real-time optimization, resulting in a more efficient optimization process.

[0161] Likewise, in Figure 3BIn step S209, the NG-RAN node 1 may obtain input data for model reasoning from the NG-RAN node 2 for optimization, wherein the input data for reasoning may include input information required from the NG-RAN node 2. Here, similarly, if the NG-RAN node 2 can execute the AI ​​ML model, the input data for reasoning may also include the corresponding reasoning results from the NG-RAN node 2. The input data for reasoning may also be associated with the same UEML context initiated for ML training. In one option, a UE ML context may be initiated before receiving the reasoning data from the NG-RAN node 2 to associate only the reasoning data of the UE ML context. For example, in the above case, the UE ML context may refer only to those UEs affected by the reasoning action.

[0162] As a result, after collecting the required input data in steps S208 and S209, model inference can then be performed in Figure 3B Here, the required measurement values ​​can also be used in model reasoning to output corresponding prediction values, including, for example, UE trajectory prediction values, target cell prediction values, target NG-RAN node prediction values, etc.

[0163] Furthermore, based on these predicted values, recommended actions or configurations, NG-RAN Node 1, NG-RAN Node 2 and the corresponding UE may again perform the corresponding AI ML based optimization / handover process to handover the UE from NG-RAN Node 1 to the target NG-RAN Node 2. Again, in one option, the optimization may be parameterized by the UE context ID.

[0164] In addition, steps S211a to S213 may be similar to Figure 3A The handover and feedback processing steps are described in steps S107a-S109.

[0165] Therefore, if the above Figure 3B As can be seen from the optimization process of the present application, the corresponding context persistence process claimed in the present application can be implemented for various different AI ML and non-AI ML optimization processes, preferably, where in particular, the UE ML context is still important even after a given handover process. To this end, it should be noted again that the configuration of the claimed method is not limited to, for example Figures 1 to 3B Rather, any of the embodiments shown in the drawings may also be varied in single or multiple process steps or equipment conditions.

[0166] For example, in a preferred embodiment of the present invention, AI ML model training does not need to be performed in the same network entity as subsequent model inference. Instead, model training can be performed in the target network entity (e.g., Figure 3B In other preferred embodiments, any of the above processing steps may also be performed in other network entities (e.g., OAM) of the corresponding RAN, so that, exemplarily, model training may be performed in OAM, while the obtained output data is further transmitted to other network entities (such as NG-RAN node 1) for model reasoning by means of output signaling. Figure 3B In another example of the present application, RAN node 2 may be replaced by OAM, so that requests may be similarly exchanged between the corresponding NG-RAN source node (NG-RAN node 1) and the OAM of the corresponding RAN.

[0167] Therefore, as a further embodiment of the present application, Figure 4 The message exchange to achieve UE ML context persistence in a general feedback-based action can also be shown, where Figure 3B Compared with the signaling process between the NG-RAN node 1 and the NG-RAN node 2, the corresponding context initiation, handover and feedback report and / or request can be performed between the corresponding NG-RAN source node (NG-RAN node 1) and the OAM of the corresponding RAN. At this time, it is considered that the construction of the corresponding signaling path between the NG-RAN source node and the OAM can be obtained from Figure 4 is derived from the disclosed content in.

[0168] The NG-RAN node 2 is assumed to optionally have an AI / ML model that can generate the required inputs, such as resource status and utilization prediction / estimation, etc. In step 301, the NG-RAN node 1 configures the measurement information on the UE side and sends a configuration message including the configuration information to the UE. In step S302, the UE collects the indicated measurement values, such as UE measurement values ​​related to RSRP, RSRQ, SINR of the serving cell and neighboring cells. In addition, in step 303, the UE sends a measurement report message including the required measurement values ​​to the NG-RAN node.

[0169] In step S304, an initiation request is provided to the target network entity for managing inter-node signaling of the UE ML context. The NG-RAN node 1 initiates the UEML context for data to be used for training the AI / ML algorithm. The NG-RAN node 1 may include an ID identifying the UE ML group and further indicate the measurements required for training the AI / ML model. These measurements need to be collected by the OAM, for example by separate signaling, when the request S304 is directly signaled to the OAM or via the NG-RAN node 2 and the input data in step S306a, as Figure 4The measurement configuration is also given to identify the requested measurement. The signaling may also include configuration information, which includes the configuration of the required measurements / counters / predicted values ​​and how these measurements should be collected (preferably, this may include an averaging window to calculate the measurement, if the measurement is based on an average, i.e., the start and end time or event, or just the start time / event and the duration of the averaging window; if a prediction is desired, the accuracy conditions that need to be met and their validity period; the measurement trigger conditions (event-based or time-based). The measurements may also include inference values ​​from the gNB or UE when making AI / ML decisions. The measurements may include UE-level performance information (e.g., bit rate, packet loss rate, latency, energy efficiency, etc.) on UEs affected by the inference action (e.g., UEs that trigger handover due to AI / ML mobility optimization decisions, or UEs connected to capacity cells that are decided to be turned off for energy saving purposes). The measurements may include cell-level performance information after the inference action is taken, such as based on multiple system KPIs (throughput, delay, RLF, etc.), counters. In step S305, the NG-RAN node 2 responds by confirming the message initiated by the UE ML context for the requested measurement value.

[0170] Therefore, a signaling-based approach is proposed, where a pre-set "UE ML Group" may correspond to a group of UEs that receive similar treatment in the RAN, e.g. due to UE history or following similar mobility paths or trajectories, radio capabilities, activated services or slices, radio conditions, UEs affected by network actions (e.g., affected by SON or AI / ML algorithm output or other network decisions), etc. A UE ML group is identified by its UE ML group index. For example, the assignment of a UE ML group may correspond to: a) all UEs that the network has selected for training ML models or algorithms (such as AI / ML energy saving, mobility optimization or load balancing, etc.), b) all UEs that the network has selected for running ML algorithm inference, etc. Each NG-RAN node will locally store information about its UE ML group in a newly introduced context "UEML Context". The network will only use UEs belonging to a UEML group to calculate counters or performance measurements based on the UE ML context.

[0171] In steps S306a and S306b, the OAM obtains input data for training the AI / ML model from each NG-RAN node, wherein the input data for training includes the required input information from these nodes. If one of the nodes executes the AI / ML model, the input data for training may include the corresponding inference results from the corresponding node. The input data received from the node may be associated with the initiated UE ML context ID. In step S307a, the training of the AI / ML model or algorithm is performed. In step S307b, the AI / ML model is deployed to the selected node, which is Figure 4 NGRAN node 1 in.

[0172] In step S308, the NG-RAN node obtains the measurement report as inference data for real-time UE optimization (e.g., mobility optimization). Subsequently, in step S309, the NG-RAN node 1 obtains input data for inference for UE optimization from the NG-RAN node 2, wherein the input data for inference includes required input information from the NG-RAN node 2. If the NG-RAN node 2 executes the AI / ML model, the input data for inference may include the corresponding inference results from the NG-RAN node 2. The input data for inference may also be associated with the same UE ML context initiated for ML training. In one option, the UE ML context may be initiated before receiving the inference data from the NG-RAN node 2 to associate the inference data only to the UE ML context. For example, in this case, the UE ML context only points to the UE affected by the inference action. Model inference is performed in step S310. Model inference is performed using the required measurement results to output prediction values, including, for example, UE trajectory prediction values, target cell prediction values, and target NG-RAN node prediction values.

[0173] In addition, steps S311a to S313 may be similar to Figure 3A The handover and feedback processing steps described in steps S107a-S109 of . Based on the predicted value, recommended action or configuration, NG-RAN node 1, the target NG-RAN node (NG-RAN node 2 is used as an example), and the UE perform an optimization action / handover process to handover the UE from NG-RAN node 1 to the target NG-RAN node. In one option, the optimization action can be parameterized by the UE ML context ID. This can inform the target gNB that performance information needs to be calculated for a given UE based on the measurement values ​​indicated in the UE ML context.

[0174] The measurement values ​​improved by ML reporting in steps S308 and / or S313 may include UE-level performance information (e.g., bit rate, packet loss rate, latency, energy efficiency, etc.) on UEs affected by the reasoning action (e.g., UEs for which handover is triggered due to AI / ML mobility optimization decisions made on them, or UEs connected to capacity cells that are decided to be shut down for energy saving purposes).

[0175] Based on these features, it is possible to store information related to the UE outside the duration that the UE is identifiable in the RAN, and to process feedback information (related to the UE ML context ID) and the network to configure the amount, trigger conditions and identification of these measurements accordingly. In addition, it is also possible to obtain UE related information from neighboring nodes outside the duration that the UE is identifiable in the RAN, at least due to the UEML group index.

[0176] Finally, it is still important to note that although in the above-described example embodiments (refer to the accompanying drawings), the messages transmitted / exchanged between network components / elements may appear to have specific / clear names, depending on various implementations (e.g., underlying technologies), these messages may have different names and / or be transmitted / exchanged in different forms / formats, which can be understood and recognized by those skilled in the art.

[0177] Although UE ML groups, UE ML group indexes and UE ML contexts are mentioned, which may be features of an AI / ML framework, in a further development, where non-AI / ML algorithms are used, surviving UE groups (or UE optimization groups), UE group indexes (or UE optimization group indexes) and UE contexts (or UE optimization contexts) may be used instead.

[0178] According to some example embodiments, corresponding methods suitable for being performed by a device (network element / component) as described above, such as a UE, CU, DU, etc., are also provided.

[0179] It should also be noted that the device (equipment) features described above correspond to corresponding method features, but for reasons of brevity, may not be explicitly described. The disclosure of this document is considered to extend to such method features as well. Specifically, the present application is understood to relate to methods of operating the above-mentioned devices, and / or providing and / or arranging individual elements of these devices.

[0180] In addition, according to some further example embodiments, a corresponding device (for example, implementing the UE, CU, DU, etc. as described above) is also provided, which includes at least one processing circuit and at least one memory for storing instructions to be executed by the processing circuit, wherein the at least one memory and the instructions are configured to, together with the at least one processing circuit, enable the corresponding device to at least perform the corresponding steps as described above.

[0181] However, in some other example embodiments, corresponding apparatuses (eg, implementing the UE, CU, DU, etc. as described above) are provided, which include corresponding apparatuses configured to at least perform the corresponding steps as described above.

[0182] It is worth noting that the examples of the embodiments of the present application are applicable to various network configurations. In other words, the examples shown in the above figures are used as the basis of the examples discussed above, which are only illustrative and do not limit the present application in any way. That is, other existing and proposed new functions available in the corresponding operating environment can be used in combination with the examples of the embodiments of the present application based on the defined principles.

[0183] It should also be noted that the disclosed example embodiments can be implemented in a variety of ways using hardware and / or software configurations. For example, the disclosed embodiments can be implemented using dedicated hardware and / or hardware in combination with software executable thereon. The components and / or elements in the accompanying drawings are provided only as examples and do not limit the scope of use or functionality of any hardware, software in combination with hardware, firmware, embedded logic components, or combinations of two or more such components that implement a particular embodiment of the present application.

[0184] It should also be noted that the description and drawings merely illustrate the principles of the present application. Those skilled in the art will be able to implement various arrangements, which, although not explicitly described or shown herein, embody the principles of the present application and are included in its spirit and scope. In addition, all examples and embodiments outlined in the present application are primarily and explicitly used only for explanatory purposes to help the reader understand the principles of the proposed method. In addition, all statements of the principles, aspects and embodiments of the present application and its specific examples provided herein are intended to cover their equivalents.

Claims

1. A device comprising: one or more processors, and a memory storing instructions that, when executed by the one or more processors, cause the apparatus to: Submitting a request by a source network entity of a radio access network RAN ​​to a target network entity, the request comprising at least a user equipment machine learning UE ML group index, wherein the UE ML group index at least indicates that the UE is part of a UE ML group; A response message is received by the source network entity from the target network entity based at least on the indicated UE ML group index.

2. The device according to claim 1, wherein: The request includes an ID for identifying a UE ML group and configuration information, which indicates the requested data collected by the target network entity at least for training or inference of a machine learning ML model.

3. The device according to any one of claims 1 or 2, wherein: The apparatus trains the ML model based at least on information of the received response message of the target network entity.

4. A device according to any one of the preceding claims, wherein: The device is caused to perform a UE handover process according to an optimization action based at least on information according to the received response message.

5. A device according to any one of the preceding claims, wherein: The configuration information further includes at least one of instructions regarding how the request data should be collected or instructions specifying timing information for the request data; The requested data includes at least one of the following: a requested measurement value, a requested counter, and a requested prediction value.

6. A device according to any one of the preceding claims, wherein: The requested measurements include at least inference information from the network node or the UE before or after taking the ML action.

7. A device according to any one of the preceding claims, wherein: The requested measurements include at least UE-level performance information on UEs affected by the inference action for which handover is triggered due to the optimization decision, or UEs connected to the capacity cell decided to be switched off.

8. A device according to any one of the preceding claims, wherein: In the use case involving UE handover, the response message is a feedback report provided after performing ML model inference.

9. The device according to any one of the preceding claims, wherein: The response message includes a UE ML group index and feedback information, where the feedback information includes information of UEs affected by the ML reasoning.

10. The device of claim 9, wherein: The feedback information includes information averaged over at least a period of time.

11. The device of claim 10, wherein: The time period includes at least a time after handover of the UE is completed at the target network entity.

12. The device according to any one of claims 9 to 11, wherein: The feedback information is recorded in an ML report, which accumulates the request data from neighboring network entities.

13. The device according to any one of the preceding claims, wherein: The request to the target network entity requests information for each UE ML group.

14. An apparatus comprising: one or more processors, and a memory storing instructions that, when executed by one or more processors, cause the apparatus to: A handover required message is transmitted by the source network entity to the target network entity, which includes a user equipment machine learning UEML group index and an instruction, wherein the instruction indicates that the feedback information should be available after the handover is completed for the UE, or the feedback information should be available after a context release message is sent from the target network entity to the source network entity.

15. The device of claim 14, wherein: The apparatus is further caused to store the information of the handoff required message at least until a release message is received at the source network entity.

16. The device according to claim 14 or 15, wherein: The apparatus causes receiving, by the source network entity, a response message from the target network entity, which is based on at least the indicated UE ML group index and includes feedback information; as well as The UE handover procedure is performed according to an optimization action based at least on the feedback information of the received response message.

17. An apparatus comprising: One or more processors and a memory storing instructions that, when executed by the one or more processors, cause the apparatus to: Receiving, at a target network entity, a request including at least a user equipment machine learning UE ML group index, wherein the UE ML group index indicates that the UE is part of a UE ML group; Based on the UE ML group index, at least collect or locally store feedback information about the UE ML group of the target network entity; Transmitting a response message to the source network entity based on the indicated UE ML group index.

18. The device of claim 17, wherein: The request is at least: a machine learning (ML) context initiation request including configuration information indicating requested data to be collected by the target network entity; or A feedback report request is used to request information of the target network entity to provide information of UEs affected by ML reasoning.

19. The device according to claim 17 or 18, wherein: Even after the UE ML context is released, the requested data is collected per UE and stored at the target network entity.

20. The device according to at least one of claims 17 to 19, wherein: In a use case involving UE handover, the response message is a feedback report provided after performing ML model inference and includes information of UEs affected by the ML inference.

21. The device according to at least one of claims 17 to 20, wherein: The apparatus is further configured to store feedback information at least until a time after handover of the UE is completed or a time after a context release message is sent from the target network entity to the source network entity.

22. The device according to at least one of claims 17 to 21, wherein: The response message includes the UE ML group index and feedback information, and the feedback information includes information averaged over at least a period of time.

23. An apparatus comprising: one or more processors, and a memory storing instructions that, when executed by the one or more processors, cause the apparatus to: In a use case including UE handover, a feedback report provided after executing a machine learning ML model action is transmitted by a target network entity of a radio access network RAN; The feedback report is transmitted when a request is received from the source network entity, or when the feedback report is available at the target network entity or at a predetermined time after handover.

24. A method of a source network entity of a Radio Access Network (RAN), the method comprising: submitting a request to a target network entity, the request comprising at least a user equipment machine learning UE ML group index, wherein the UE ML group index indicates that the UE is part of a UE ML group; A response message is received from the target network entity based at least on the indicated UE ML group index.

25. A method for a target network entity of a Radio Access Network (RAN), the method comprising: receiving a request from a source network entity, the request comprising at least a user equipment machine learning UE ML group index, wherein the UE ML group index indicates that the UE is part of a UE ML group; Based on the UE ML group index, at least collect or locally store feedback information about the UE ML group of the target network entity; Transmitting a response message to the source network entity based on the indicated UE ML group index.

26. A computer program product comprising a set of instructions which, when executed on a device, are configured to cause the device to perform the method as claimed in any one of claims 24 or 25.

27. The computer program product of claim 26, embodied as a computer readable medium or directly loadable into a computer.

28. An apparatus comprising: One or more processors and a memory storing instructions that, when executed by one or more processors, cause the apparatus to: Submitting a request by a source network entity of a radio access network RAN ​​to a target network entity, the request comprising at least a user equipment UE optimization group index, wherein the UE optimization group index at least indicates that the UE is part of a UE optimization group; A response message is received by the source network entity from the target network entity based at least on the indicated UE optimization group index.

29. An apparatus comprising: One or more processors and a memory storing instructions that, when executed by one or more processors, cause the apparatus to: receiving, at a target network entity, a request including at least a user equipment UE optimization group index, wherein the UE optimization group index indicates that the UE is part of a UE optimization group; Based on the UE optimization group index, at least collect or locally store feedback information about the UE optimization group of the target network entity; A response message is transmitted to the source network entity based on the indicated UE optimization group index.

Images