Devices, methods, apparatuses, and computer readable media for fallback of machine learning functionality

Abstract

Disclosed are devices, methods, apparatuses, and computer readable media for fallback of machine learning functionality An example apparatus for a terminal device may include at least one processor and at least one memory. The at least one memory may store instructions that, when executed by the at least one processor, may cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; receive from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, apply modifications to the first configuration for use during fallback, and enable the second configuration in the network function.

Description

DEVICES, METHODS, APPARATUSES, AND COMPUTER READABLE MEDIA OR FALLBACK OF MACHINE LEARNING FUNCTIONALITYTECHNICAL FIELD

[0001] Various example embodiments relate to devices, methods, apparatuses, and computer readable media for fallback of machine learning functionality.BACKGROUND

[0002] Artificial intelligence (Al) / machine learning (ML) -enabled feature refers to a feature where AI / ML may be used. AI / ML is a topic in most 3rd Generation Partnership Project (3GPP) standardization work groups. An aim is to enable the support of AI / ML functionalities through the use of AI / ML models. The AI / ML models, once deployed into the functions of the cellular communication system, could potentially have a direct impact on the behavior of the system. The capability and performance of the ML model are determined by every operational phase of its lifecycle: training, testing, deployment, and inference phases. The AI / ML functionalities using these operational phases may be applied to both infrastructure network elements and terminal devices (user equipment). Functionality may refer to an AI / ML-enabled Feature / Feature Group (FG) enabled by configuration(s), where configuration(s) is(are) supported based on conditions indicated by user equipment (UE) capability. Correspondingly, functionality-based life cycle management (LCM) operates based on, at least, one configuration of AI / ML-enabled Feature / FG or specific configurations of an AI / ML-enabled Feature / FG.

[0003] For AI / ML enhancements related to beam management (BM), two sub-use cases have been identified: beam prediction in the spatial domain, which may be referred to as BM-Casel, and beam prediction in the time domain, which may be referred to as BM-Case2. The scope of spatial beam prediction (BM-Casel) is to predict the best downlink (DL) transmit (Tx) beam and / or DL Tx / receive (Rx) beam pairs in different spatial locations. The scope of time-domain beam prediction (BM-Case2) aims to predict the best DL Tx beam and / or DL Tx / Rx beam pairs to use for next time instant. DL Tx beam prediction for both UE-sided model and network (NW)- sided model encompasses spatial-domain DL Tx beam prediction for Set A of beams based on measurement results of Set B of beams (“BM-Casel”) and temporal DL Tx beam prediction forSet A of beams based on the historic measurement results of Set B of beams (“BM-Case2”).SUMMARY

[0004] A brief summary of exemplary embodiments is provided below to provide basic understanding of some aspects of various embodiments. It should be noted that this summary is not intended to identify key features of essential elements or define scopes of the embodiments, and its sole purpose is to introduce some concepts in a simplified form as a preamble for a more detailed description provided below.

[0005] In a first aspect, disclosed is an apparatus for a terminal device. The apparatus may include at least one processor and at least one memory. The at least one memory may store instructions that, when executed by the at least one processor, may cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; receive from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, apply modifications to the first configuration for use during fallback, and enable the second configuration in the network function.

[0006] In a second aspect, disclosed is an apparatus for a network device. The apparatus may include at least one processor and at least one memory. The at least one memory may store instructions that, when executed by the at least one processor, may cause the apparatus at least to: transmit to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; determine performance of the machine learning functionality; and transmit to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

[0007] In a third aspect, disclosed is an apparatus for a terminal device. The apparatus may include at least one processor and at least one memory. The at least one memory may storeinstructions that, when executed by the at least one processor, may cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; determine performance of the machine learning functionality; and transmit to the network, a first request for activating fallback from the machine learning functionality in the network function based on the determined performance of the machine learning functionality.

[0008] In a fourth aspect, disclosed is an apparatus for a network device. The apparatus may include at least one processor and at least one memory. The at least one memory may store instructions that, when executed by the at least one processor, may cause the apparatus at least to: transmit to a terminal device, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; receive from the terminal device, a first request for activating fallback from the machine learning functionality in the network function; and in response to the first request, transmit to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality.

[0009] In a fifth aspect, disclosed is a method performed by an apparatus for a terminal device. The method may comprise: receiving from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; receiving from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, applying modifications to the first configuration for use during fallback, and enabling the second configuration in the network function.

[0010] In a sixth aspect, disclosed is a method performed by an apparatus for a network device. The method may comprise: transmitting to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the secondconfiguration is a fallback configuration from the first configuration; determining performance of the machine learning functionality; and transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

[0011] In a seventh aspect, disclosed is a method performed by an apparatus for a terminal device. The method may comprise: receiving from a network, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a nonmachine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; determining performance of the machine learning functionality; and transmitting to the network, a first request for activating fallback from the machine learning functionality in the network function based on the determined performance of the machine learning functionality.

[0012] In an eighth aspect, disclosed is a method performed by an apparatus for a network device. The method may comprise: transmitting to a terminal device, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; receiving from the terminal device, a first request for activating fallback from the machine learning functionality in the network function; and in response to the first request, transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality.

[0013] In a ninth aspect, disclosed is an apparatus for a terminal device. The apparatus may comprise: means for receiving from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a nonmachine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; means for receiving from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and means for in response to the first indication, applying modifications to the first configuration for use during fallback, and enabling the second configuration the second configuration in the network function.

[0014] In a tenth aspect, disclosed is an apparatus for a network device. The apparatus may comprise: means for transmitting to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; means for determining performance of the machine learning functionality; and means for transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

[0015] In an eleventh aspect, disclosed is an apparatus for a terminal device. The apparatus may comprise: means for receiving from a network, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; means for determining performance of the machine learning functionality; and means for transmitting to the network, a first request for activating fallback from the machine learning functionality in the network function based on the determined performance of the machine learning functionality.

[0016] In a twelfth aspect, disclosed is an apparatus for a network device. The apparatus may comprise: means for transmitting to a terminal device, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non- machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; means for receiving from the terminal device, a first request for activating fallback from the machine learning functionality in the network function; and means for in response to the first request, transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality.

[0017] In a thirteenth aspect, a computer-readable medium is disclosed. The computer-readable medium may comprise program instructions that, when executed by an apparatus for a terminal device, may cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function,wherein the second configuration is a fallback configuration from the first configuration; receive from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, apply modifications to the first configuration for use during fallback, and enable the second configuration the second configuration in the network function.

[0018] In a fourteenth aspect, a computer-readable medium is disclosed. The computer- readable medium may comprise program instructions that, when executed by an apparatus for a network device, may cause the apparatus at least to: transmit to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; determine performance of the machine learning functionality; and transmit to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

[0019] In a fifteenth aspect, a computer-readable medium is disclosed. The computer-readable medium may comprise program instructions that, when executed by an apparatus for a terminal device, may cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; determine performance of the machine learning functionality; and transmit to the network, a first request for activating fallback from the machine learning functionality in the network function based on the determined performance of the machine learning functionality.

[0020] In a sixteenth aspect, a computer-readable medium is disclosed. The computer-readable medium may comprise program instructions that, when executed by an apparatus for a network device, may cause the apparatus at least to: transmit to a terminal device, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; receive from theterminal device, a first request for activating fallback from the machine learning functionality in the network function; and in response to the first request, transmit to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality.

[0021] Other features and advantages of the example embodiments of the present disclosure will also be apparent from the following description of specific embodiments when read in conjunction with the accompanying drawings, which illustrate, by way of example, the principles of example embodiments of the present disclosure.BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Some example embodiments will now be described, by way of non-limiting examples, with reference to the accompanying drawings.

[0023] FIG. 1 shows an example sequence diagram according to the example embodiments of the present disclosure.

[0024] FIG. 2 shows a flow chart illustrating an example method 200 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0025] FIG. 3 shows a flow chart illustrating an example method 300 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0026] FIG. 4 shows a flow chart illustrating an example method 400 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0027] FIG. 5 shows a flow chart illustrating an example method 500 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0028] FIG. 6 shows a block diagram illustrating an example device 600 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0029] FIG. 7 shows a block diagram illustrating an example device 700 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0030] FIG. 8 shows a block diagram illustrating an example apparatus 800 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0031] FIG. 9 shows a block diagram illustrating an example apparatus 900 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0032] FIG. 10 shows a block diagram illustrating an example apparatus 1000 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0033] FIG. 11 shows a block diagram illustrating an example apparatus 1100 for fallback of machine learning functionality according to the example embodiments of the present disclosure.

[0034] Throughout the drawings, same or similar reference numbers indicate same or similar elements. A repetitive description on the same elements would be omitted.DETAILED DESCRIPTION

[0035] Herein below, some example embodiments are described in detail with reference to the accompanying drawings. The following description includes specific details for the purpose of providing a thorough understanding of various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. In some instances, well known circuits, techniques and components are shown in block diagram form to avoid obscuring the described concepts and features.

[0036] Non-AI / ML features are assumed to be available for operation when an AI / ML functionality becomes non-applicable. So, a term “fallback” was introduced as an operation to configure the UE with “to non-AI / ML operation (i.e., not relying on inference process)”. Such definition of fallback implies that a UE previously configured with an AI / ML functionality would have the AI / ML functionality deactivated and have a non-AI / ML features activated. The NW would be in charge of the procedure to reconfigure the UE. In the present disclosure, the term “AI / ML functionality” can be interchangeable with a term “machine learning functionality,” following the literature in the art.

[0037] Currently if an AI / ML functionality performs poorly, it is assumed that the UE is unable to switch to a different model to improve performance for the configured AI / ML functionality, thus the UE should be reconfigured with another AI / ML functionality or with a fallback mechanism based on non-AI / ML features. However, in some cases, for example, the model may not generalize well and they would be trained using data collected from specific areas, cell sites, scenarios, etc., the poor performance of an AI / ML functionality could be temporary, a temporary impairment may be resolved, and the model could return to good performance.

[0038] Example embodiments of the present disclosure provide solutions for a temporaryfallback of a machine learning functionality, by re-enabling the use of an AI / ML functionality which performed well for some time in the past but it is experiencing poor performance temporarily and can ensure good performance again after temporary poor performance, at least as good as the non-AI / ML feature. According to the example embodiments of the present disclosure, an AI / ML functionality temporary fallback mechanism can enable the UE to keep the AI / ML functionality configuration and assist either UE or NW on performance monitoring of AI / ML functionality without a need for deactivate and activate the configuration again. In other words, for the AI / ML functionality whose poor performance is temporary, a reconfiguration for another AI / ML functionality can be avoided.

[0039] FIG. 1 shows an example sequence diagram according to the example embodiments of the present disclosure. Referring to FIG. 1, a UE 110 may represent any terminal device in a network, a network device 150 may represent the network side serving the UE 110. The network device 150 may be a base station (BS), such as an Evolved Node B (eNB), a next Generation Node B (gNB), etc. From the perspective of the UE 110, the network device 150 may be referred to as a network.

[0040] It is assumed that capability exchange, general radio resource control (RRC) configuration and reconfiguration mechanisms, and other mature mechanisms have already occurred between the UE 110 and the network device 150. It is also assumed that AI / ML applicable functionality has been indicated. In the present disclosure, beam management and channel state information (CSI) feedback enhancement are taken as examples of AI / ML functionality use cases which are supported by interactions between the UE 110 and the network device 150. Such use cases or applications relevant to the use cases such as CSI compression, CSI prediction, temporal beam prediction, spatial beam prediction are taken as examples of network functions. Use of AI / ML functionalities for such network functions have been readily disclosed in the literature and, since the present disclosure is not focused on the details of the training or inference execution, such disclosure is omitted from the present document. Those skilled in the art may understand that the example embodiments of the present disclosure may also apply in other AI / ML functionality use cases such as positioning enhancement, etc.

[0041] The network device 150 may transmit to the UE 110 at least one first configuration 152 for a machine learning functionality of a determined network function, e.g. the beammanagement or CSI prediction, and a second configuration 154 for a non-machine learning functionality of the determined network function.

[0042] The first configuration 152 and the second configuration 154 may be transmitted in a single message or separate messages. In some embodiments, the message carrying the first configuration 152 and / or the second configuration 154 may be RRC message(s), for example, RRC reconfiguration message, such as CSI-ReportConfig(s) defined in 3GPP technical specification (TS) 38.331.

[0043] The first configuration 152 and the second configuration 154 serve the identical network function. In other words, the machine learning functionality and the non-machine learning functionality target / are used for the same network function / application or use case. The second configuration 154 may be a fallback configuration from the first configuration 152. In some embodiments, the second configuration 154 may comprise a non-ML configuration, e.g., the non- machine learning functionality parameters. In fallback mode, when the second configuration 154 is enabled, a fallback from the ML functionality occurs but the first configuration 152 is still active.

[0044] In some embodiments, the first configuration 152 may comprise machine learning functionality parameters, the machine learning functionality parameters may be normal parameters for configuring the machine learning functionality. In some embodiments, the first configuration 152 may comprise the modifications to be applied to the machine learning parameters during fallback. In some embodiments, the first configuration 152 may comprise a pointer pointing to the second configuration 154. In some embodiments, the second configuration 154 may be received / transmitted as an identifier or an index, and the pointer in the first configuration 152 may point to the identifier or the index of the second configuration 154.

[0045] In some embodiments, the machine learning functionality parameters may be modified by the UE 110 during fallback, and the modified machine learning functionality parameters may be termed as fallback related parameters. In some embodiments, in the fallback mode, the fallback related parameters may replace at least part of the normal parameters in the machine learning functionality, for example, to enable a lower reporting frequency. The non-ML functionality may perform inference, report, or other operations up to its own fallback related parameters.

[0046] In some embodiments, in the fallback mode, the first configuration 152 can be used for making decision to exit fallback or to permanently fall back to the non-machine learning functionality. In some embodiments, the second configuration 154 may be the fallback configuration to which the pointer of the first configuration 152 points. In some embodiments, in the fallback mode, the second configuration 154 can be used by the network to make decisions, for example, on beam selection.

[0047] In the present disclosure, “during fallback” and “in fallback mode” can be interchangeable, following the literature in the art.

[0048] In some embodiments, during fallback, the non-ML configuration can be enabled and used by the network to make decisions. For example, during fallback the non-ML configuration can be used to generate reports for beam management in a non-ML way.

[0049] In some embodiments, the first configuration 152 may be indicated by CS ReportConfigld = 1, and the second configuration 154 may be indicated by CSI-ReportConfigld = O.Then, the network device 150 may activate the first configuration 152 CSI-ReportConfig=l. Thus, the UE 110 is configured to use one or more AI / ML models to perform inference, enabling the AI / ML functionality. Depending on the configured type of report, the UE 110 may then report to the network device 150 the results of the particular AI / ML functionality and the network function, such as beam predictions or CSI predictions.

[0050] In some embodiments, in an operation 112, the UE 110 may monitor the performance of the machine learning functionality, and then in an operation 114, the UE 110 may determine the performance of the machine learning functionality based on monitoring the performance of the machine learning functionality performed in the operation 112. Then, in some embodiments, the UE 110 may transmit to the network device 150, a first request 118 for activating fallback from the machine learning functionality in the network function based on the determined performance of the machine learning functionality performed in the operation 114.

[0051] For example, if the UE 110 in the operation 112 executes performance monitoring for the machine learning functionality and in the operation 114 determines that the performance has dropped to an unacceptable level, e.g., a degradation in the performance of the machine learning functionality, the UE 110 may send to the network device 150 the first request 118 for a temporary fallback from the machine learning functionality, indicating the CSI-ReportConfigldfrom which to temporarily fallback, and in this case CSI-ReportConfigld = 1.

[0052] In response to the first request 118, the network device 150 may transmit to the UE 110, a first indication 160 indicating the UE 110 to activate fallback from the machine learning functionality.

[0053] In some embodiments, in an operation 158, the network device 150 may determine the performance of the machine learning functionality. In some embodiments, in an operation 156, the network device 150 may monitor the performance of the machine learning functionality, and then in the operation 158, the network device 150 may determine the performance of the machine learning functionality based on the monitoring of the performance of the machine learning functionality performed in the operation 156. Alternatively or additionally, in some embodiments, after the operation 112, the UE 110 may report to the network device 150, information 116 on the performance of the machine learning functionality, and then in the operation 158, the network device 150 may determine the performance of the machine learning functionality based on the information 116 reported from the UE 110 on the performance of the machine learning functionality. Thus, in some embodiments, in the operation 158, the network device 150 may determine the performance of the machine learning functionality based on at least one of the following: monitoring of the performance of the machine learning functionality performed in the operation 156; or the information 116 reported from the UE 110 on the performance of the machine learning functionality. In this case, the network device 150 may transmit to the UE 110, the first indication 160 indicating the UE 110 to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality performed in the operation 158.

[0054] For example, if the network device 150 in the operation 156 executes performance monitoring for the machine learning functionality and in the operation 158 determines that the performance has dropped to an unacceptable level, e.g., a degradation in the performance of the machine learning functionality, the network device 150 may send to the UE 110 the first indication 160 commanding the UE 110 to activate a temporary fallback from the machine learning functionality, CSI-ReportConfigld = 1.

[0055] If the UE 110 reports to the network device 150, the information 116 on the performance of the machine learning functionality and in the operation 158, the network device 150determines the performance of the machine learning functionality based on the information 116 reported from the UE 110 on the performance of the machine learning functionality, e.g., a degradation in the performance of the machine learning functionality, the UE 110 may receive the first indication 160 based on the reporting of the information 116.

[0056] In response to the first indication 160, in an operation 120, the UE 110 may activate fallback in the network function. In some embodiments, in the operation 120, the UE 110 may apply modifications to the first configuration 152 for use during fallback, and enable the second configuration 154. For example, the UE 110 may activate the second configuration 154, CSI- ReportConfigld = 0. In the fallback mode, the UE 110 may apply the fallback-related parameters. Then, the UE 110 may report to the network device 150 the beam management or CSI feedback measurements, depending on the use case.

[0057] In some embodiments, in fallback mode, the UE 110 and / or the network device 150 may use outputs of the machine learning functionality for monitoring the machine learning functionality, and the UE 110 and / or the network device 150 may not use the machine learning functionality for making decisions for the network function.

[0058] In some embodiments, in the fallback mode, the second configuration 154 is enabled, and during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0059] In some embodiments, the second configuration 154 may consume less resources compared with the first configuration 152, in terms of at least one of the following: lower reporting frequency; fewer measurements to report; or more limited reference signal (RS) resources to measure. For example, in fallback mode, the UE 110 could be configured to report AI / ML-based outputs at a lower frequency such that the network device 150 would be able to evaluate the performance without impacting overall network performance during the temporary fallback.

[0060] In fallback mode, for the termination of the temporary fallback, either the UE 110 or the network device 150 may determine performance of the AI / ML functionality from which fallback is activated.

[0061] In some embodiments, in an operation 122, the UE 110 may monitor the performance of the machine learning functionality, and then in an operation 124, the UE 110 may determinethe performance of the machine learning functionality based on monitoring the performance of the machine learning functionality performed in the operation 122. Then, in some embodiments, the UE 110 may transmit to the network device 150, a second request 128 for exiting fallback in the network function based on the determined performance of the machine learning functionality performed in the operation 124.

[0062] For example, if the UE 110 in the operation 122 executes performance monitoring for the machine learning functionality based on a configured parameter and in the operation 124 determines that the performance has risen to an acceptable level, e.g., an improvement in the performance of the machine learning functionality, the UE 110 may send to the network device 150 the second request 128 for exiting / releasing the temporary fallback, and in this case CSI- ReportConfigld = 1.

[0063] In response to the second request 128, the network device 150 may transmit to the UE 110, a second indication 166 indicating the UE 110 to exit fallback.

[0064] In some embodiments, in an operation 164, the network device 150 may determine the performance of the machine learning functionality. In some embodiments, in an operation 162, the network device 150 may monitor the performance of the machine learning functionality, and then in the operation 164, the network device 150 may determine the performance of the machine learning functionality based on the monitoring of the performance of the machine learning functionality performed in the operation 162. Alternatively or additionally, in some embodiments, after the operation 122, the UE 110 may report to the network device 150, information 126 on the performance of the machine learning functionality, and then in the operation 164, the network device 150 may determine the performance of the machine learning functionality based on the information 126 reported from the UE 110 on the performance of the machine learning functionality. Thus, in some embodiments, in the operation 164, the network device 150 may determine the performance of the machine learning functionality based on at least one of the following: monitoring of the performance of the machine learning functionality performed in the operation 162; or the information 126 reported from the UE 110 on the performance of the machine learning functionality. In this case, the network device 150 may transmit to the UE 110, the second indication 166 indicating the UE 110 to exit fallback based on the determined performance of the machine learning functionality performed in the operation 164.

[0065] For example, if the network device 150 in the operation 162 executes performance monitoring for the machine learning functionality and in the operation 164 determines that the performance has risen to an acceptable level, e.g., an improvement in the performance of the machine learning functionality, the network device 150 may send to the UE 110 the second indication 166 commanding the UE 110 to exit fallback, CSI-ReportConfigld = 1.

[0066] In response to the second indication 166, in an operation 130, the UE 110 may undo the modifications to the first configuration 152 for use during fallback and switch disable from the second configuration 154 to exit fallback.

[0067] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function. In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used by the network device 150 and / or the UE 110 for making decisions for the network function.

[0068] For example, after the operation 130, beam predictions or CSI predictions would be reported using the non-fallback configuration and non-AI / ML measurement reports would be disabled.

[0069] In some embodiments, compared to the first configuration 152 or the second configuration 154, the first indication 160 or the second indication 166 may be transmitted / received via a lower layer message. For example, the first configuration 152 or the second configuration 154 may be carried by an RRC message, e.g., CSI-ReportConfig, and the first indication 160 or the second indication 166 may be carried by a downlink control information (DCI) command, a medium access control (MAC) control element (MAC-CE) command, or a combination of MAC-CE and DCI commands, for triggering entering or exiting / releasing a CSI-reporting associated temporary fallback.

[0070] In some embodiments, in case the second configuration 154 is received / transmitted as its identifier or index, the first indication 160 and / or the second indication 166 may carry the pointer and / or the identifier or the index to indicate the activating / entering or exiting / releasing of the fallback regarding the second configuration 154. Since the pointer in the first configuration 152 points to the second configuration 154, based on the pointer and / or the identifier or the index,the UE 110 may determine the correct second configuration 154 and corresponding parameters.

[0071] Thus, the example embodiments of the present disclosure provide low latency mechanisms for the UE 110 to activate / enter or exit / release fallback mode for a specific and activated AI / ML functionality.

[0072] The fallback related parameters in the example embodiments of the present disclosure are described below.

[0073] For the beam management and CSI prediction use cases, the main parameter required to enable temporary fallback may comprise the pointer to the fallback CSI report configuration, which may be denoted as reportConfigForFallbackld.

[0074] The following additional parameters could be configured to define the behavior of the AI / ML functionality during temporary fallback in a less resource intensive manner than having the UE measure RS and transmit reports simultaneously for an AI / ML functionality and a non- ML feature.

[0075] Report configuration for fallback type, which may be denoted as reportConfigForFallbackType, would configure the UE with a different beam prediction or CSI prediction reporting periodicity than when not configured in temporary fallback. For example, the purpose would be to assist network side performance monitoring of the AI / ML configuration.

[0076] This monitoring configuration may be reversed from the usual monitoring configuration. Usually, the latter, a non-AI / ML beam management feature would be enabled for occasional reporting to calculate the performance metric of the AI / ML functionality. In this monitoring configuration, the former, the AI / ML beam management functionality is being configured for occasional reporting, while the non-AI / ML BM feature is being configured for full-time reporting to take over for the AI / ML functionality.

[0077] Resources for channel measurement for fallback, which may be denoted as resourcesForChannelMeasurementForFallback, could configure the UE with different or more limited RS resources to measure while the AI / ML configuration is in temporary fallback mode.

[0078] Report quantity for fallback, which may be denoted as reportQuantityForFallback, could configure the UE to disable reporting using the setting “none”. During this time, as long as the UE could still measure the CSI resources used for the AI / ML configuration, the UE would be able to execute its own performance monitoring for the AI / ML configuration.

[0079] As an alternative, the UE could instead be configured to report measurements of the following: synchronization signal block or synchronization signal and physical broadcast channel (PBCH) block (SSB) beams resources corresponding to Set B beams (in case of wide- beams to narrow-beam prediction alternative) which can be known as fallback to Pl; and / or CSI resources corresponding to Set B beams (in case of Set B beams are subset of Set A beams).

[0080] The fallback versions of existing parameters may directly replace the existing parameters with the fallback versions, if configured, during temporary fallback mode.

[0081] According to the example embodiments of the present disclosure, an example representing the possible impact of the protocol message on a standard, e.g. 3GPP TS 38.331, may be shown as Table 1 below, where some relevant excerpts of the full message are included, and removed text is denoted by

[0082] Table 1 CSI-ReportConfig Enhancements- ASN1 START- TAG-CSI-REPORTCONFIG- STARTCSI-ReportConfig ::= SEQUENCE ! reportConfigld CSI-ReportConfigld, resourcesF orChannelMeasurement CSI-ResourceConfigld, reportConfigType CHOICE { periodic SEQUENCE { reportSlotConfig CSI-ReportPeriodicityAndOffset, semiPer si stentOnPU C CH SEQUENCE { semiPer si stentOnPU S CH SEQUENCE { aperiodic SEQUENCE {}, reportQuantity CHOICE { none NULL, cri-RSRP NULL, ssb-Index-RSRP NULL,},[[ reportConfigForFallbackId-rl9 CSI-ReportConfigld OPTIONAL, resourceForChannelMeasurementForFallback-rl9 CSI-ResourceConfigld OPTIONAL, reportConfigForFallbackType-rl9 CHOICE {[The same options could be used as in reportConfigType]} OPTIONAL, reportQuantityForFallback-rl9 CHOICE {[The same options could be used as in reportQuantity]} OPTIONAL

[0083] Example MAC-CE definitions to enter and exit temporary fallback mode according to the example embodiments of the present disclosure are described below. The beam management and CSI prediction use cases, in which functionalities would be defined using CSI report configurations, identified by CSI report configuration identifiers, denoted as CSL ReportConfigld, can be used for illustration. For each example, assume that a field in the general MAC header identifies the CE type. The MAC CEs would be transmitted on the downlink from the network to the UE as a command, e.g. the first indication 160 and / or the second indication 166 or on the uplink from the UE to the network as a request, e.g. the first request 118 and / or the second request 128. “maxNoCSI-ReportConfigld” in the tables below represent the maximum number of CSI-ReportConfigld.

[0084] Table 2 shows an example MAC-CE to toggle temporary fallback mode.

[0085] Table 2

[0086] In the MAC-CE shown in Table 2, the only field other than its identifier in the overall MAC header is the CSI-ReportConfigld.

[0087] Table 3 shows an example MAC-CE to enter or exit temporary fallback mode

[0088] Table 3

[0089] The MAC-CE shown in Table 3 adds a field with an explicit action to enter or exitfallback mode.

[0090] Table 4 shows an example MAC-CE to enter or exit temporary fallback mode or to permanently fallback to non-AI / ML

[0091] Table 4

[0092] The MAC-CE shown in Table 4 adds a field to enable the same MAC-CE to support permanent or temporary fallback mode.

[0093] The example embodiments of the present disclosure enable the network to provide a simple command which simultaneously enables a non-AI / ML feature and an AI / ML functionality in a mode that allows the network to monitor the performance of the AI / ML functionality (which is not in use). Once the performance of the AI / ML functionality improves, the network can make the decision to exit fallback mode, which would deactivate the non-AI / ML feature, and reactivate the AI / ML functionality with its previous configuration.

[0094] FIG. 2 shows a flow chart illustrating an example method 200 for fallback of machine learning functionality according to the example embodiments of the present disclosure. The example method 200 may be performed, for example, by an apparatus for a terminal device, such as the UE 110 above mentioned.

[0095] Referring to FIG. 2, the example method 200 may comprise: an operation 210 of receiving from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; an operation 220 of receiving from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and an operation 230 of in response to the first indication, applying modifications to the first configuration for use during fallback, and enabling the second configuration the second configuration in the network function.

[0096] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parametersduring fallback, and a pointer pointing to the second configuration, and wherein the second configuration may comprise non-machine learning functionality parameters.

[0097] In some embodiments, the example method 200 may comprise: during fallback, using outputs of the machine learning functionality for monitoring the machine learning functionality and not using the machine learning functionality for making decisions for the network function.

[0098] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0099] In some embodiments, the example method 200 may comprise: receiving from the network, a second indication indicating the terminal device to exit fallback; and in response to the second indication, undoing the modifications to the first configuration for use during fallback and disabling the second configuration to exit fallback.

[0100] In some embodiments, the exiting from fallback comprises at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0101] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0102] In some embodiments, the first indication or the second indication may be received via a lower layer message compared to the first configuration or the second configuration.

[0103] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: lower reporting frequency; fewer measurements to report; or more limited reference signal resources to measure.

[0104] In some embodiments, the example method 200 may comprise: monitoring performance of the machine learning functionality; reporting to the network, information on the performance of the machine learning functionality; and receiving the first indication based on the reporting.

[0105] In some embodiments, the second configuration may be received as an identifier or an index.

[0106] FIG. 3 shows a flow chart illustrating an example method 300 for fallback of machine learning functionality according to the example embodiments of the present disclosure. Theexample method 300 may be performed, for example, by an apparatus for a network device, such as the network device 150 above mentioned.

[0107] Referring to FIG. 3, the example method 300 may comprise: an operation 310 of transmitting to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; an operation 320 of determining performance of the machine learning functionality; and an operation 330 of transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

[0108] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may non-machine learning functionality parameters.

[0109] In some embodiments, the example method 300 may comprise: during fallback, using outputs of the machine learning functionality for monitoring the machine learning functionality, and not using the machine learning functionality for making decisions for the network function.

[0110] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0111] In some embodiments, the example method 300 may comprise: transmitting to the terminal device, a second indication indicating the terminal device to exit fallback based on the determined performance of the machine learning functionality.

[0112] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0113] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0114] In some embodiments, the first indication or the second indication may be transmittedvia a lower layer message compared to the first configuration or the second configuration.

[0115] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: lower reporting frequency; fewer measurements to report; or more limited reference signal resources to measure.

[0116] In some embodiments, the example method 300 may comprise: determining the performance of the machine learning functionality based on at least one of the following: monitoring the performance of the machine learning functionality; or information, reported from the terminal device, on the performance of the machine learning functionality.

[0117] In some embodiments, the second configuration may be transmitted as an identifier or an index.

[0118] FIG. 4 shows a flow chart illustrating an example method 400 for fallback of machine learning functionality according to the example embodiments of the present disclosure. The example method 400 may be performed, for example, by an apparatus for a terminal device, such as the UE 110 above mentioned.

[0119] Referring to FIG. 4, the example method 400 may comprise: an operation 410 of receiving from a network, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; an operation 420 of determining performance of the machine learning functionality; and an operation 430 of transmitting to the network, a first request for activating fallback from the machine learning functionality in the network function based on the determined performance of the machine learning functionality.

[0120] In some embodiments, the example method 400 may comprise: in response to the first request, receiving from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, applying modifications to the first configuration for use during fallback, and enabling the second configuration in the network function.

[0121] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the secondconfiguration may comprise non-machine learning functionality parameters.

[0122] In some embodiments, the example method 400 may comprise: during fallback, using outputs of the machine learning functionality for monitoring the machine learning functionality and not using the machine learning functionality for making decisions for the network function.

[0123] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0124] In some embodiments, the example method 400 may comprise: transmitting to the network, a second request for exiting from fallback in the network function based on the determined performance of the machine learning functionality; in response to a second request, receiving from the network, a second indication indicating the terminal device to exit fallback; and in response to the second indication, disabling the second configuration and undoing the modifications to the first configuration for use during fallback to exit fallback.

[0125] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0126] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0127] In some embodiments, the first indication or the second indication may be received via a lower layer message compared to the first configuration or the second configuration.

[0128] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: fewer measurements to monitor; or more limited reference signal resources to measure.

[0129] In some embodiments, the example method 400 may comprise: determining the performance of the machine learning functionality based on monitoring the performance of the machine learning functionality.

[0130] In some embodiments, the second configuration may be received as an identifier or an index.

[0131] FIG. 5 shows a flow chart illustrating an example method 500 for fallback of machinelearning functionality according to the example embodiments of the present disclosure. The example method 500 may be performed, for example, by an apparatus for a network device, such as the network device 150 above mentioned.

[0132] Referring to FIG. 5, the example method 500 may comprise: an operation 510 of transmitting to a terminal device, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; an operation 520 of receiving from the terminal device, a first request for activating fallback from the machine learning functionality in the network function; and an operation 530 of in response to the first request, transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality.

[0133] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may comprise non-machine learning functionality parameters.

[0134] In some embodiments, the example method 500 may comprise: during fallback, using outputs of the machine learning functionality for monitoring the machine learning functionality and not using the machine learning functionality for making decisions for the network function.

[0135] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0136] In some embodiments, the example method 500 may comprise: receiving from the terminal device, a second request for exiting from fallback in the network function; and in response to the second request, transmitting to the terminal device, a second indication indicating the terminal device to exit fallback.

[0137] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0138] In some embodiments, during the enabled inference of the machine learningfunctionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0139] In some embodiments, the first indication or the second indication may be transmitted via a lower layer message compared to the first configuration or the second configuration.

[0140] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: fewer measurements to monitor; or more limited reference signal resources to measure.

[0141] In some embodiments, the second configuration may be transmitted as an identifier or an index.

[0142] FIG. 6 shows a block diagram illustrating an example device 600 for fallback of machine learning functionality according to the example embodiments of the present disclosure. The device, for example, may be at least part of an apparatus for a terminal device, such as the UE 110 in the above examples.

[0143] As shown in FIG. 6, the example device 600 may include at least one processor 610 and at least one memory 620 that may store instructions 630. The instructions 630, when executed by the at least one processor 610, may cause the device 600 at least to perform the example method 200 or 400 described above.

[0144] In various example embodiments, the at least one processor 610 in the example device 600 may include, but is not limited to, at least one hardware processor, including at least one microprocessor such as a central processing unit (CPU), a portion of at least one hardware processor, and any other suitable dedicated processor such as those developed based on for example Field Programmable Gate Array (FPGA) and Application Specific Integrated Circuit (ASIC). Further, the at least one processor 610 may also include at least one other circuitry or element not shown in FIG. 6.

[0145] In various example embodiments, the at least one memory 620 in the example device 600 may include at least one storage medium in various forms, such as a transitory memory and / or a non-transitory memory. The transitory memory may include, but is not limited to, for example, a random-access memory (RAM), a cache, and so on. The non-transitory memory may include, but is not limited to, for example, a read-only memory (ROM), a hard disk, a flash memory, and so on. The term “non-transitory,” as used herein, is a limitation of the medium itself(i.e., tangible, not a signal) as opposed to a limitation on data storage persistency (e.g., RAM vs. ROM). Further, the at least memory 620 may include, but is not limited to, an electric, a magnetic, an optical, an electromagnetic, an infrared, or a semiconductor system, apparatus, or device or any combination of the above.

[0146] Further, in various example embodiments, the example device 600 may also include at least one other circuitry, element, and interface, for example at least one I / O interface, at least one antenna element, and the like.

[0147] In various example embodiments, the circuitries, parts, elements, and interfaces in the example device 600, including the at least one processor 610 and the at least one memory 620, may be coupled together via any suitable connections including, but is not limited to, buses, crossbars, wiring and / or wireless lines, in any suitable ways, for example electrically, magnetically, optically, electromagnetically, and the like.

[0148] It is understood that the structure of the device on the side of the UE 110 is not limited to the above example device 600.

[0149] FIG. 7 shows a block diagram illustrating an example device 700 for fallback of machine learning functionality according to the example embodiments of the present disclosure. The device, for example, may be at least part of an apparatus for a network device, such as the network device 150 in the above examples.

[0150] As shown in FIG. 7, the example device 700 may include at least one processor 710 and at least one memory 720 that may store instructions 730. The instructions 730, when executed by the at least one processor 710, may cause the device 700 at least to perform the example method 300 or 500 described above.

[0151] In various example embodiments, the at least one processor 710 in the example device 700 may include, but is not limited to, at least one hardware processor, including at least one microprocessor such as a central processing unit (CPU), a portion of at least one hardware processor, and any other suitable dedicated processor such as those developed based on for example Field Programmable Gate Array (FPGA) and Application Specific Integrated Circuit (ASIC). Further, the at least one processor 710 may also include at least one other circuitry or element not shown in FIG. 7.

[0152] In various example embodiments, the at least one memory 720 in the example device700 may include at least one storage medium in various forms, such as a transitory memory and / or a non-transitory memory. The transitory memory may include, but is not limited to, for example, a random-access memory (RAM), a cache, and so on. The non-transitory memory may include, but is not limited to, for example, a read-only memory (ROM), a hard disk, a flash memory, and so on. The term “non-transitory,” as used herein, is a limitation of the medium itself (i.e., tangible, not a signal) as opposed to a limitation on data storage persistency (e.g., RAM vs. ROM). Further, the at least memory 720 may include, but is not limited to, an electric, a magnetic, an optical, an electromagnetic, an infrared, or a semiconductor system, apparatus, or device or any combination of the above.

[0153] Further, in various example embodiments, the example device 700 may also include at least one other circuitry, element, and interface, for example at least one I / O interface, at least one antenna element, and the like.

[0154] In various example embodiments, the circuitries, parts, elements, and interfaces in the example device 700, including the at least one processor 710 and the at least one memory 720, may be coupled together via any suitable connections including, but is not limited to, buses, crossbars, wiring and / or wireless lines, in any suitable ways, for example electrically, magnetically, optically, electromagnetically, and the like.

[0155] It is understood that the structure of the device on the side of the network device 150 is not limited to the above example device 700.

[0156] FIG. 8 shows a block diagram illustrating an example apparatus 800 for fallback of machine learning functionality according to the example embodiments of the present disclosure. The apparatus, for example, may be at least part of a terminal device, such as the UE 110 in the above examples.

[0157] As shown in FIG. 8, the example apparatus 800 may comprise: means 810 for receiving from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; means 820 for receiving from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and means 830 for in response to the first indication, applying modifications to thefirst configuration for use during fallback, and enabling the second configuration in the network function.

[0158] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may comprise non-machine learning functionality parameters.

[0159] In some embodiments, the apparatus 800 may comprise: means for during fallback, using outputs of the machine learning functionality for monitoring the machine learning functionality and not using the machine learning functionality for making decisions for the network function.

[0160] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0161] In some embodiments, the apparatus 800 may comprise: means for receiving from the network, a second indication indicating the terminal device to exit fallback; and means for in response to the second indication, undoing the modifications to the first configuration for use during fallback and disabling the second configuration to exit fallback.

[0162] In some embodiments, the exiting from fallback comprises at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0163] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0164] In some embodiments, the first indication or the second indication may be received via a lower layer message compared to the first configuration or the second configuration.

[0165] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: lower reporting frequency; fewer measurements to report; or more limited reference signal resources to measure.

[0166] In some embodiments, the apparatus 800 may comprise: means for monitoring performance of the machine learning functionality; means for reporting to the network,information on the performance of the machine learning functionality; and means for receiving the first indication based on the reporting.

[0167] In some embodiments, the second configuration may be received as an identifier or an index.

[0168] In some example embodiments, examples of means in the example apparatus 800 may include circuitries. For example, an example of means 810 may include a circuitry configured to perform the operation 210 of the example method 200, an example of means 820 may include a circuitry configured to perform the operation 220 of the example method 200, and an example of means 830 may include a circuitry configured to perform the operation 230 of the example method 200.

[0169] The example apparatus 800 may further include means comprising circuitry configured to perform the example method 200. In some example embodiments, examples of means may also include software modules and any other suitable function entities.

[0170] FIG. 9 shows a block diagram illustrating an example apparatus 900 for fallback of machine learning functionality according to the example embodiments of the present disclosure. The apparatus, for example, may be at least part of a network device, such as the network device 150 in the above examples.

[0171] As shown in FIG. 9, the example apparatus 900 may comprise: means 910 for transmitting to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; means 920 for determining performance of the machine learning functionality; and means 930 for transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

[0172] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may comprise non-machine learning functionality parameters.

[0173] In some embodiments, the apparatus 900 may comprise: means for during fallback,using outputs of the machine learning functionality for monitoring the machine learning functionality and not using the machine learning functionality for making decisions for the network function.

[0174] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0175] In some embodiments, the apparatus 900 may comprise: means for transmitting to the terminal device, a second indication indicating the terminal device to exit fallback based on the determined performance of the machine learning functionality.

[0176] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0177] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0178] In some embodiments, the first indication or the second indication may be transmitted via a lower layer message compared to the first configuration or the second configuration.

[0179] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: lower reporting frequency; fewer measurements to report; or more limited reference signal resources to measure.

[0180] In some embodiments, the apparatus 900 may comprise: means for determining the performance of the machine learning functionality based on at least one of the following: monitoring the performance of the machine learning functionality; or information, reported from the terminal device, on the performance of the machine learning functionality.

[0181] In some embodiments, the second configuration may be transmitted as an identifier or an index.

[0182] In some example embodiments, examples of means in the example apparatus 900 may include circuitries. For example, an example of means 910 may include a circuitry configured to perform the operation 310 of the example method 300, an example of means 920 may include a circuitry configured to perform the operation 320 of the example method 300, and an exampleof means 930 may include a circuitry configured to perform the operation 330 of the example method 300.

[0183] The example apparatus 900 may further include means comprising circuitry configured to perform the example method 300. In some example embodiments, examples of means may also include software modules and any other suitable function entities.

[0184] FIG. 10 shows a block diagram illustrating an example apparatus 1000 for fallback of machine learning functionality according to the example embodiments of the present disclosure. The apparatus, for example, may be at least part of a terminal device, such as the UE 110 in the above examples.

[0185] As shown in FIG. 10, the example apparatus 1000 may comprise: means 1010 for receiving from a network, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; means 1020 for determining performance of the machine learning functionality; and means 1030 for transmitting to the network, a first request for activating fallback from the machine learning functionality in the network function based on the determined performance of the machine learning functionality.

[0186] In some embodiments, the apparatus 1000 may comprise: means for in response to the first request, receiving from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and means for in response to the first indication, applying modifications to the first configuration for use during fallback, and enabling the second configuration in the network function..

[0187] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may non-machine learning functionality parameters.

[0188] In some embodiments, the apparatus 1000 may comprise: means for during fallback, using outputs of the machine learning functionality for monitoring the machine learning functionality and not using the machine learning functionality for making decisions for the network function.

[0189] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0190] In some embodiments, the apparatus 1000 may comprise: means for transmitting to the network, a second request for exiting from fallback in the network function based on the determined performance of the machine learning functionality; means for in response to a second request, receiving from the network, a second indication indicating the terminal device to exit fallback; and means for in response to the second indication, disabling the second configuration and undoing the modifications to the first configuration for use during fallback to exit fallback.

[0191] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0192] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0193] In some embodiments, the first indication or the second indication may be received via a lower layer message compared to the first configuration or the second configuration.

[0194] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: fewer measurements to monitor; or more limited reference signal resources to measure.

[0195] In some embodiments, the apparatus 1000 may comprise: means for determining the performance of the machine learning functionality based on monitoring the performance of the machine learning functionality.

[0196] In some embodiments, the second configuration may be received as an identifier or an index.

[0197] In some example embodiments, examples of means in the example apparatus 1000 may include circuitries. For example, an example of means 1010 may include a circuitry configured to perform the operation 410 of the example method 400, an example of means 1020 may include a circuitry configured to perform the operation 420 of the example method 400, and an example of means 1030 may include a circuitry configured to perform the operation 430 of the examplemethod 400.

[0198] The example apparatus 1000 may further include means comprising circuitry configured to perform the example method 400. In some example embodiments, examples of means may also include software modules and any other suitable function entities.

[0199] FIG. 11 shows a block diagram illustrating an example apparatus 1100 for fallback of machine learning functionality according to the example embodiments of the present disclosure. The apparatus, for example, may be at least part of a network device, such as the network device 150 in the above examples.

[0200] As shown in FIG. 11, the example apparatus 1100 may comprise: means 1110 for transmitting to a terminal device, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; means 1120 for receiving from the terminal device, a first request for activating fallback from the machine learning functionality in the network function; and means 1130 for in response to the first request, transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality.

[0201] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may comprise non-machine learning functionality parameters.

[0202] In some embodiments, the apparatus 1100 may comprise: means for during fallback, using outputs of the machine learning functionality for monitoring the machine learning functionality and not using the machine learning functionality for making decisions for the network function.

[0203] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0204] In some embodiments, the apparatus 1100 may comprise: means for receiving from the terminal device, a second request for exiting from fallback in the network function; and meansfor in response to the second request, transmitting to the terminal device, a second indication indicating the terminal device to exit fallback.

[0205] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0206] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0207] In some embodiments, the first indication or the second indication may be transmitted via a lower layer message compared to the first configuration or the second configuration.

[0208] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: fewer measurements to monitor; or more limited reference signal resources to measure.

[0209] In some embodiments, the second configuration may be transmitted as an identifier or an index.

[0210] In some example embodiments, examples of means in the example apparatus 1100 may include circuitries. For example, an example of means 1110 may include a circuitry configured to perform the operation 510 of the example method 500, an example of means 1120 may include a circuitry configured to perform the operation 520 of the example method 500, and an example of means 1130 may include a circuitry configured to perform the operation 530 of the example method 500.

[0211] The example apparatus 1100 may further include means comprising circuitry configured to perform the example method 500. In some example embodiments, examples of means may also include software modules and any other suitable function entities.

[0212] The example embodiments of the present disclosure also provide a computer-readable medium comprising program instructions that, when executed by an apparatus for a terminal device, such as the UE 110 in the above examples, may cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallbackconfiguration from the first configuration; receive from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, apply modifications to the first configuration for use during fallback, and enable the second configuration in the network function.

[0213] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may non-machine learning functionality parameters.

[0214] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: during fallback, use outputs of the machine learning functionality for monitoring the machine learning functionality and not use the machine learning functionality for making decisions for the network function.

[0215] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0216] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: receive from the network, a second indication indicating the terminal device to exit fallback; and in response to the second indication, ndo the modifications to the first configuration for use during fallback and disable the second configuration to exit fallback.

[0217] In some embodiments, the exiting from fallback comprises at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0218] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0219] In some embodiments, the first indication or the second indication may be received via a lower layer message compared to the first configuration or the second configuration.

[0220] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: lower reporting frequency;fewer measurements to report; or more limited reference signal resources to measure.

[0221] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: monitor performance of the machine learning functionality; report to the network, information on the performance of the machine learning functionality; and receive the first indication based on the reporting.

[0222] In some embodiments, the second configuration may be received as an identifier or an index.

[0223] The example embodiments of the present disclosure also provide a computer-readable medium comprising program instructions that, when executed by an apparatus for a network device, such as the network device 150 in the above examples, may cause the apparatus at least to: transmit to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; determine performance of the machine learning functionality; and transmit to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

[0224] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may comprise non-machine learning functionality parameters.

[0225] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: during fallback, use outputs of the machine learning functionality for monitoring the machine learning functionality, and not use the machine learning functionality for making decisions for the network function.

[0226] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0227] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: transmit to the terminal device, asecond indication indicating the terminal device to exit fallback based on the determined performance of the machine learning functionality.

[0228] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0229] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0230] In some embodiments, the first indication or the second indication may be transmitted via a lower layer message compared to the first configuration or the second configuration.

[0231] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: lower reporting frequency; fewer measurements to report; or more limited reference signal resources to measure.

[0232] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: determine the performance of the machine learning functionality based on at least one of the following: monitoring the performance of the machine learning functionality; or information, reported from the terminal device, on the performance of the machine learning functionality.

[0233] In some embodiments, the second configuration may be transmitted as an identifier or an index.

[0234] The example embodiments of the present disclosure also provide a computer-readable medium comprising program instructions that, when executed by an apparatus for a terminal device, such as the UE 110 in the above examples, may cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; determine performance of the machine learning functionality; and transmit to the network, a first request for activating fallback from the machine learning functionality in the network function based on the determined performance of the machine learning functionality.

[0235] In some embodiments, the computer-readable medium may include instructions that,when executed by the apparatus, may cause the apparatus to: in response to the first request, receive from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, apply modifications to the first configuration for use during fallback, and enable the second configuration in the network function.

[0236] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may comprise non-machine learning functionality parameters.

[0237] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: during fallback, use outputs of the machine learning functionality for monitoring the machine learning functionality and not use the machine learning functionality for making decisions for the network function.

[0238] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the network function.

[0239] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: transmit to the network, a second request for exiting from fallback in the network function based on the determined performance of the machine learning functionality; in response to a second request, receive from the network, a second indication indicating the terminal device to exit fallback; and in response to the second indication, disable the second configuration and undo the modifications to the first configuration for use during fallback to exit fallback.

[0240] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0241] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0242] In some embodiments, the first indication or the second indication may be received viaa lower layer message compared to the first configuration or the second configuration.

[0243] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: fewer measurements to monitor; or more limited reference signal resources to measure.

[0244] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: determine the performance of the machine learning functionality based on monitoring the performance of the machine learning functionality.

[0245] In some embodiments, the second configuration may be received as an identifier or an index.

[0246] The example embodiments of the present disclosure also provide a computer-readable medium comprising program instructions that, when executed by an apparatus for a network device, such as the network device 150 in the above examples, may cause the apparatus at least to: transmit to a terminal device, at least one first configuration for a machine learning functionality for a network function, and a second configuration for a non-machine learning functionality for the network function, wherein the second configuration is a fallback configuration from the first configuration; receive from the terminal device, a first request for activating fallback from the machine learning functionality in the network function; and in response to the first request, transmit to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality.

[0247] In some embodiments, the first configuration may comprise machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration may comprise non-machine learning functionality parameters.

[0248] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: during fallback, use outputs of the machine learning functionality for monitoring the machine learning functionality and not use the machine learning functionality for making decisions for the network function.

[0249] In some embodiments, during the enabled non-machine learning functionality, outputs of the non-machine learning functionality may be used for decision making in the networkfunction.

[0250] In some embodiments, the computer-readable medium may include instructions that, when executed by the apparatus, may cause the apparatus to: receive from the terminal device, a second request for exiting from fallback in the network function; and in response to the second request, transmit to the terminal device, a second indication indicating the terminal device to exit fallback.

[0251] In some embodiments, the exiting from fallback may comprise at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

[0252] In some embodiments, during the enabled inference of the machine learning functionality, outputs of the machine learning functionality may be used for making decisions for the network function.

[0253] In some embodiments, the first indication or the second indication may be transmitted via a lower layer message compared to the first configuration or the second configuration.

[0254] In some embodiments, the second configuration may consume less resources compared with the first configuration, in terms of at least one of the following: fewer measurements to monitor; or more limited reference signal resources to measure.

[0255] In some embodiments, the second configuration may be transmitted as an identifier or an index.

[0256] As used herein, “at least one of the following: ” and “at least one of ” and similar wording, where the list of two or more elements are joined by “and” or “or”, mean at least any one of the elements, or at least any two or more of the elements, or at least all the elements.

[0257] The term “terminal device” refers to any end device that may be capable of wireless communication. By way of example rather than limitation, a terminal device may also be referred to as a communication device, user equipment (UE), a Subscriber Station (SS), a Portable Subscriber Station, a Mobile Station (MS), or an Access Terminal (AT). The terminal device may include, but is not limited to, a mobile phone, a cellular phone, a smart phone, voice over IP (VoIP) phones, wireless local loop phones, a tablet, a wearable terminal device, a personal digital assistant (PDA), portable computers, desktop computer, image capture terminal devices such asdigital cameras, gaming terminal devices, music storage and playback appliances, vehiclemounted wireless terminal devices, wireless endpoints, mobile stations, laptop-embedded equipment (LEE), laptop-mounted equipment (LME), USB dongles, smart devices, wireless customer-premises equipment (CPE), an Internet of Things (loT) device, a watch or other wearable, a head-mounted display (HMD), a vehicle, a drone, a medical device and applications (e.g., remote surgery), an industrial device and applications (e.g., a robot and / or other wireless devices operating in an industrial and / or an automated processing chain contexts), a consumer electronics device, a device operating on commercial and / or industrial wireless networks, and the like. The terminal device may also correspond to a Mobile Termination (MT) part of an IAB node (e.g., a relay node). In the above description, the terms “terminal device”, “communication device”, “terminal”, “user equipment” and “UE” may be used interchangeably.

[0258] The term “circuitry” throughout this disclosure may refer to one or more or all of the following: (a) hardware-only circuit implementations (such as implementations in only analog and / or digital circuitry); (b) combinations of hardware circuits and software, such as (as applicable) (i) a combination of analog and / or digital hardware circuit(s) with software / firmware and (ii) any portions of hardware processor(s) with software (including digital signal processor(s)), software, and memory(ies) that work together to cause an apparatus, such as a mobile phone or server, to perform various functions); and (c) hardware circuit(s) and or processor(s), such as a microprocessor(s) or a portion of a microprocessor(s), that requires software (e.g., firmware) for operation, but the software may not be present when it is not needed for operation. This definition of circuitry applies to one or all uses of this term in this disclosure, including in any claims. As a further example, as used in this disclosure, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. The term circuitry also covers, for example and if applicable to the claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in server, a cellular network device, or other computing or network device.

[0259] Another example embodiment may relate to computer program codes or instructions which may cause an apparatus to perform at least the respective methods described above. Another example embodiment may be related to a computer-readable medium having suchcomputer program codes or instructions stored thereon. In some embodiments, such a computer- readable medium may include at least one storage medium in various forms such as a volatile memory and / or a non-volatile memory. The volatile memory may include, but is not limited to, for example, a RAM, a cache, and so on. The non-volatile memory may include, but is not limited to, a ROM, a hard disk, a flash memory, and so on. The non-volatile memory may also include, but is not limited to, an electric, a magnetic, an optical, an electromagnetic, an infrared, or a semiconductor system, apparatus, or device or any combination of the above.

[0260] Unless the context clearly requires otherwise, throughout the description and the claims, the words “comprise,” “comprising,” and the like are to be construed in an inclusive sense, as opposed to an exclusive or exhaustive sense; that is to say, in the sense of “including, but is not limited to.” The word “coupled”, as generally used herein, refers to two or more elements that may be either directly connected, or connected by way of one or more intermediate elements. Likewise, the word “connected”, as generally used herein, refers to two or more elements that may be either directly connected, or connected by way of one or more intermediate elements. Additionally, the words “herein,” “above,” “below,” and words of similar import, when used in this application, shall refer to this application as a whole and not to any particular portions of this application. Where the context permits, words in the description using the singular or plural number may also include the plural or singular number respectively. The word “or” in reference to a list of two or more items, that word covers all of the following interpretations of the word: any of the items in the list, all of the items in the list, and any combination of the items in the list.

[0261] Moreover, conditional language used herein, such as, among others, “can,” “could,” “might,” “may,” “e.g.,” “for example,” “such as” and the like, unless specifically stated otherwise, or otherwise understood within the context as used, is generally intended to convey that certain embodiments include, while other embodiments do not include, certain features, elements and / or states. Thus, such conditional language is not generally intended to imply that features, elements and / or states are in any way required for one or more embodiments or that one or more embodiments necessarily include logic for deciding, with or without author input or prompting, whether these features, elements and / or states are included or are to be performed in any particular embodiment.

[0262] As used herein, the term "determine / determining" (and grammatical variants thereof)can include, not least: calculating, computing, processing, deriving, measuring, investigating, looking up (for example, looking up in a table, a database or another data structure), ascertaining and the like. Also, "determining" can include receiving (for example, receiving information), accessing (for example, accessing data in a memory), obtaining and the like. Also, "determine / determining" can include resolving, selecting, choosing, establishing, and the like.

[0263] While some embodiments have been described, these embodiments have been presented by way of example, and are not intended to limit the scope of the disclosure. Indeed, the apparatus, methods, and systems described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions, and changes in the form of the methods and systems described herein may be made without departing from the spirit of the disclosure. For example, while blocks are presented in a given arrangement, alternative embodiments may perform similar functionalities with different components and / or circuit topologies, and some blocks may be deleted, moved, added, subdivided, combined, and / or modified. At least one of these blocks may be implemented in a variety of different ways. The order of these blocks may also be changed. Any suitable combination of the elements and actions of the some embodiments described above can be combined to provide further embodiments. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the disclosure.

[0264] Abbreviations used in the description and / or in the figures are defined as follows: 3GPP 3rd Generation Partnership ProjectAl artificial intelligenceBS base stationBM beam managementDCI downlink control informationDL downlink eNB Evolved Node BFG Feature Group gNB next Generation Node BLCM life cycle managementMAC medium access controlMAC CE MAC control elementML machine learningNW networkRRC radio resource controlRS reference signalCSI channel state informationRx receivePBCH physical broadcast channelSSB synchronization signal block synchronization signal and PBCH blockTS technical specificationTx transmitUE user equipment

Claims

WHAT IS CLAIMED IS:

1. An apparatus for a terminal device, comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; receive from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, apply modifications to the first configuration for use during fallback, and enable the second configuration in the network function.

2. The apparatus of claim 1, wherein the first configuration comprises machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration comprises non-machine learning functionality parameters.

3. The apparatus of claim 2, wherein the apparatus is configured to, during fallback, use outputs of the machine learning functionality for monitoring the machine learning functionality and not use the machine learning functionality for making decisions for the network function.

4. The apparatus of claim 2 or 3, wherein during the enabled non-machine learning functionality, outputs of the non-machine learning functionality are used for decision making in the network function.

5. The apparatus of any of claims 1 to 4, wherein the apparatus is configured to: receive from the network, a second indication indicating the terminal device to exit fallback;45and in response to the second indication, undo the modifications to the first configuration for use during fallback and disable the second configuration to exit fallback.

6. The apparatus of claim 5, wherein the exiting from fallback comprises at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

7. The apparatus of claim 6, wherein during the enabled inference of the machine learning functionality, outputs of the machine learning functionality are used for making decisions for the network function.

8. The apparatus of any of claims 5 to 7, wherein the first indication or the second indication is received via a lower layer message compared to the first configuration or the second configuration.

9. The apparatus of any of claims 1 to 8, wherein the second configuration consumes less resources compared with the first configuration, in terms of at least one of the following: lower reporting frequency; fewer measurements to report; or more limited reference signal resources to measure.

10. The apparatus of any of claims 1 to 9, wherein the apparatus is configured to: monitor performance of the machine learning functionality; report to the network, information on the performance of the machine learning functionality; and receive the first indication based on the reporting.

11. The apparatus of any of claims 1 to 10, wherein the second configuration is received as46an identifier or an index.

12. An apparatus for a network device, comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: transmit to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; determine performance of the machine learning functionality; and transmit to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

13. The apparatus of claim 12, wherein the first configuration comprises machine learning functionality parameters, the modifications to be applied to the machine learning parameters during fallback, and a pointer pointing to the second configuration, and wherein the second configuration comprises non-machine learning functionality parameters.

14. The apparatus of claim 13, wherein the apparatus is configured to, during fallback, use outputs of the machine learning functionality for monitoring the machine learning functionality and not use the machine learning functionality for making decisions for the network function.

15. The apparatus of claim 13 or 14, wherein during the enabled non-machine learning functionality, outputs of the non-machine learning functionality are used for decision making in the network function.

16. The apparatus of any of claims 12 to 15, wherein the apparatus is configured to: transmit to the terminal device, a second indication indicating the terminal device to exit47fallback based on the determined performance of the machine learning functionality.

17. The apparatus of claim 16, wherein the exiting from fallback comprises at least one of the following: enabling inference of the machine learning functionality for the network function; and disabling the non-machine learning functionality for the network function.

18. The apparatus of claim 17, wherein during the enabled inference of the machine learning functionality, outputs of the machine learning functionality are used for making decisions for the network function.

19. The apparatus of any of claims 16 to 18, wherein the first indication or the second indication is transmitted via a lower layer message compared to the first configuration or the second configuration.

20. The apparatus of any of claims 12 to 19, wherein the second configuration consumes less resources compared with the first configuration, in terms of at least one of the following: lower reporting frequency; fewer measurements to report; or more limited reference signal resources to measure.

21. The apparatus of any of claims 12 to 20, wherein the apparatus is configured to: determine the performance of the machine learning functionality based on at least one of the following: monitoring the performance of the machine learning functionality; or information, reported from the terminal device, on the performance of the machine learning functionality.

22. The apparatus of any of claims 12 to 21, wherein the second configuration is transmitted as an identifier or an index.

23. A method performed by an apparatus for a terminal device, comprising: receiving from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; receiving from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, applying modifications to the first configuration for use during fallback, and enabling the second configuration in the network function.

24. A method performed by an apparatus for a network device, comprising: transmitting to a terminal device, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; determining performance of the machine learning functionality; and transmitting to the terminal device, a first indication indicating the terminal device to activate fallback from the machine learning functionality based on the determined performance of the machine learning functionality.

Images