Methods and apparatus for artificial intelligence model management in mobile communications
Patent Information
- Application Number
- PCT/CN2026/086505
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-03-28
- Filing Date
- 2026-03-27
- Publication Date
- 2026-10-01
Smart Images

Figure CN2026086505_01102026_PF_FP_ABST
Abstract
Description
METHODS AND APPARATUS FOR ARTIFICIAL INTELLIGENCE MODEL MANAGEMENT IN MOBILE COMMUNICATIONSCROSS REFERENCE TO RELATED PATENT APPLICATION (S)
[0001] The present disclosure is part of a non-provisional application claiming the priority benefit of PCT Application No. PCT / CN2025 / 085808, filed 28 March 2025, the content of which is herein incorporated by reference in its entirety.TECHNICAL FIELD
[0002] The present disclosure is generally related to mobile communications and, more particularly, to artificial intelligence (AI) model management with respect to an apparatus and a network node in mobile communications. BACKGROUN
[0003] Unless otherwise indicated herein, approaches described in this section are not prior art to the claims listed below and are not admitted as prior art by inclusion in this section.
[0004] Wireless communication systems may be widely deployed to provide various telecommunication services such as telephony, video, data, messaging, and broadcasts. Typical wireless communication systems may use multiple-access technologies capable of supporting communication with multiple users by sharing available system resources. Examples of such multiple-access technologies may include code division multiple access (CDMA) systems, time division multiple access (TDMA) systems, frequency division multiple access (FDMA) systems, orthogonal frequency division multiple access (OFDMA) systems, single-carrier frequency division multiple access (SC-FDMA) systems, and time division synchronous code division multiple access (TD-SCDMA) systems.
[0005] In communication technologies, artificial intelligence (AI) and machine learning (ML) have permeated a wide spectrum of industries, ushering in substantial productivity enhancements. In the mobile communications systems, the technologies are orchestrating transformative shifts. The mobile devices may be progressively supplanting conventional algorithms with AI or ML models.
[0006] One challenge in applying AI for mobile wireless communication is maintaining appropriate AI models when considering the mobility of the device. Due to the limitations of the AI model generation, the performance of the AI model will decrease when devices move outside the suitable region of the AI model. Therefore, a method to monitor and manage AI models to make sure the AI model is not out-of-date or not suitable is necessary.
[0007] The general functional framework of management procedure may currently be under discussion. However, the details of the signaling and the mechanism of the AI model management procedure are still unclear. The detailed description of how to perform AI model management and the necessary modification of a specific use case, e.g., an AI model, is not clear.
[0008] Accordingly, how to perform the AI model management becomes an important issue for the newly developed wireless communication network. Therefore, there is a need to provide proper schemes for the AI model management.SUMMARY
[0009] The following summary is illustrative only and is not intended to be limiting in any way. That is, the following summary is provided to introduce concepts, highlights, benefits, and advantages of the novel and non-obvious techniques described herein. Select implementations are further described below in the detailed description. Thus, the following summary is not intended to identify essential features of the claimed subject matter, nor is it intended for use in determining the scope of the claimed subject matter.
[0010] One objective of the present disclosure is to propose schemes, concepts, designs, systems, methods, and apparatus pertaining to artificial intelligence (AI) model management with respect to an apparatus and a network node in mobile communications. It is believed that the above-described issue would be avoided or otherwise alleviated by implementing one or more of the proposed schemes described herein.
[0011] In one aspect, a method may involve an apparatus receiving a configuration from a network node. The method may also involve the apparatus performing a measurement according to the configuration to generate inference data and monitoring data for an AI inference and a model monitoring. The method may further involve the apparatus determining an operation for AI model management according to a decision result for the model monitoring.
[0012] In another aspect, a method may involve a network node transmitting a configuration for collecting inference data and monitoring data for an AI inference and a model monitoring to a user equipment (UE) . The method may also involve the network node determining an operation for AI model management according to a decision result for the model monitoring.
[0013] It is noteworthy that, although description provided herein may be in the context of certain radio access technologies, networks and network topologies such as 5th Generation System (5GS) and 4G EPS mobile networking, the proposed concepts, schemes and any variation (s) / derivative (s) thereof may be implemented in, for and by other types of wireless and wired communication technologies, networks and network topologies such as, for example and without limitation, Ethernet, Universal Terrestrial Radio Access Network (UTRAN) , E-UTRAN, Global System for Mobile communications (GSM) , General Packet Radio Service (GPRS) / Enhanced Data rates for Global Evolution (EDGE) Radio Access Network (GERAN) , Long-Term Evolution (LTE) , LTE-Advanced, LTE-Advanced Pro, IoT, Industrial IoT (IIoT) , Narrow Band Internet of Things (NB-IoT) , 6th Generation (6G) , and any future-developed networking technologies. Thus, the scope of the present disclosure is not limited to the examples described herein.BRIEF DESCRIPTION OF THE DRAWINGS
[0014] The accompanying drawings are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of the present disclosure. The drawings illustrate implementations of the disclosure and, together with the description, serve to explain the principles of the disclosure. It is appreciable that the drawings are not necessarily in scale as some components may be shown to be out of proportion than the size in actual implementation in order to clearly illustrate the concept of the present disclosure.
[0015] FIG. 1 is a diagram depicting an example scenario of a communication environment in which various solutions and schemes in accordance with the present disclosure may be implemented.
[0016] FIG. 2 is a diagram depicting an example scenario for an AI model management process under a first proposed scheme for AI model management in accordance with implementations of the present disclosure.
[0017] FIG. 3 is a diagram depicting an example scenario for an AI model management process under a second proposed scheme for AI model management in accordance with implementations of the present disclosure.
[0018] FIG. 4 is a diagram depicting an example scenario for an AI model management process under a third proposed scheme for AI model management in accordance with implementations of the present disclosure.
[0019] FIG. 5 is a diagram depicting an example scenario for an AI model management process under a fourth proposed scheme for AI model management in accordance with implementations of the present disclosure.
[0020] FIG. 6 is a diagram depicting an example scenario for an inference data configuration and a monitor data configuration in accordance with implementations of the present disclosure.
[0021] FIG. 7 is a diagram depicting an example scenario for an inter-frequency prediction in accordance with implementations of the present disclosure.
[0022] FIG. 8 is a block diagram of an example communication system in accordance with an implementation of the present disclosure.
[0023] FIG. 9 is a flowchart of an example process in accordance with an implementation of the present disclosure.
[0024] FIG. 10 is a flowchart of an example process in accordance with another implementation of the present disclosure. DETAILED DESCRIPTION OF PREFERRED IMPLEMENTATIONS
[0025] Detailed embodiments and implementations of the claimed subject matters are disclosed herein. However, it shall be understood that the disclosed embodiments and implementations are merely illustrative of the claimed subject matters which may be embodied in various forms. The present disclosure may, however, be embodied in many different forms and should not be construed as limited to the exemplary embodiments and implementations set forth herein. Rather, these exemplary embodiments and implementations are provided so that description of the present disclosure is thorough and complete and will fully convey the scope of the present disclosure to those skilled in the art. In the description below, details of well-known features and techniques may be omitted to avoid unnecessarily obscuring the presented embodiments and implementations. Overview
[0026] Implementations in accordance with the present disclosure relate to various techniques, methods, schemes, and / or solutions pertaining to artificial intelligence (AI) model management with respect to user equipment (UE) and network apparatus in mobile communications. According to the present disclosure, a number of possible solutions may be implemented separately or jointly. That is, although these possible solutions may be described below separately, two or more of these possible solutions may be implemented in one combination or another.
[0027] FIG. 1 illustrates an example scenario 100 of a communication environment in which various solutions and schemes in accordance with the present disclosure may be implemented. Scenario 100 involves a UE 110 in wireless communication with a network 120 (e.g., a wireless network including an NTN and a TN) via a terrestrial network node 125 (e.g., an evolved Node-B (eNB) , a Next Generation Node-B (gNB) , or a transmission / reception point (TRP) ) and / or a non-terrestrial network node 128 (e.g., a satellite) . For example, the terrestrial network node 125 and / or the non-terrestrial network node 128 may form a non-terrestrial network (NTN) serving cell for wireless communication with the UE 110. In some implementations, the UE 110 may be an IoT device such as an NB-IoT UE or an enhanced machine-type communication (eMTC) UE (e.g., a bandwidth reduced low complexity (BL) UE or a coverage enhancement (CE) UE) . In such a communication environment, the UE 110, the network 120, the terrestrial network node 125, and the non-terrestrial network node 128 may implement various schemes pertaining to improved AI model management procedures in accordance with the present disclosure, as described below. It is noteworthy that, while the various proposed schemes may be individually or separately described below, in actual implementations, some or all of the proposed schemes may be utilized or otherwise implemented jointly. Of course, each of the proposed schemes may be utilized or otherwise implemented individually or separately.
[0028] According to the implementations of the present disclosure, an apparatus (e.g., UE 110) may receive a configuration from a network node (e.g., terrestrial network node 125) . Then, the apparatus may perform a measurement according to the configuration to generate inference data (or model input or measure data) and monitoring data for an AI inference and a model monitoring. Then, the apparatus may determine an operation for AI model management according to a decision result for the model monitoring. According to the implementations of the present disclosure, the operation may comprise at least one of a fallback operation, a deactivate operation, a switch operation, and keeping inference (i.e., no action) .
[0029] According to the implementations of the present disclosure, the AI model management procedure may comprise configuration, data collection, AI inference, model monitoring, decision (or management decision) , and report. The management procedure can be categorized (e.g., the first proposed scheme, the second proposed scheme, the third proposed scheme, and the fourth proposed scheme of the present disclosure) based on where the operations of the AI inference, the model monitoring, and the decision are performed on the UE side or the NW side.
[0030] According to the implementations of the present disclosure, the configuration may comprise at least one of an inference configuration, a monitor configuration, and a monitor criteria configuration.
[0031] According to the implementations of the present disclosure, the inference configuration (or inference data configuration) may comprise at least one of information of an observed target and information of a predicted target.
[0032] The apparatus may measure the required data based on the information of the observed target. The corresponding measurement may be used as the input of the AI model for AI inference. The information of the observed target may comprise at least one of the target measurement object (MO) , the location (e.g., frequency and time) of the observed target, reference signal (e.g., synchronization signal block (SSB) , channel state information-reference signal (CSI-RS) , etc. ) , and the type of measurement (e.g., layer 1 (L1) beam reference signal received power (RSRP) , L1 cell RSRP, layer 3 (L3) beam RSRP, L3 cell RSRP, signal-to-interference plus noise ratio (SINR) , reference signal received quality (RSRQ) , etc. ) .
[0033] In addition, the information of the predicted target may indicate the target of the AI inference output. The information of the predicted target may comprise at least one of the predicted target MO, the target cell identifier (ID) of the predicted target, the location (e.g., frequency and time) of the predicted target, the reference signal, and the type of prediction. According to the configuration of inference data (i.e., inference configuration) , the apparatus can obtain the AI model input and output, and the necessary information for collecting the inference data for the AI inference.
[0034] According to the implementations of the present disclosure, the monitor configuration may comprise information of a ground-truth target. Specifically, the monitor configuration (or monitoring data configuration) may comprise the information for measuring the ground-truth target. The measurement of the ground-truth target may be used to examine the performance of the AI inference. In one implementation, the ground-truth target may be the same as the predicted target given in the inference data configuration with less frequency. In another embodiment, the ground-truth target may be different from the predicted target. The information of the ground-truth target may comprise at least one of the measurement object, the cell ID, the location, the reference signal, and the measurement type. According to the monitoring data configuration, the apparatus can measure the necessary data to perform the model monitoring.
[0035] The monitoring data may be configured as the results generated from the non-AI approach. The results may be the real measurement and / or the information derived from the real measurement. In the implementation, the apparatus may be configured to run two parallel procedures for the same task, e.g., one procedure is the AI approach and the other procedure is the legacy non-AI approach. In an implementation, the AI approach may be the AI mobility, and the legacy non-AI approach may be the L3 handover (HO) , lower layer triggered mobility (LTM) , conditional handover (CHO) , etc.
[0036] According to the implementations of the present disclosure, the monitor criteria configuration may comprise an L1 or L3 cell RSRP difference.
[0037] Under a first proposed scheme for AI model management, the apparatus may perform the AI inference according to the inference data (or model input or measure data) to generate an inference result. Then, the apparatus may perform the model monitoring according to the inference result and the monitoring data to generate a model monitoring result. Then, the apparatus may determine the decision result according to the model monitoring result. Then, the apparatus may transmit a decision report to the network node according to the decision result. Specifically, according to the first proposed scheme for AI model management, the apparatus may perform the AI inference, the model monitoring, and the management decision. In the first proposed scheme, the network node may configure the inference configuration, the monitor configuration, and the monitor criteria configuration for the apparatus. According to the inference configuration, the monitor configuration, and the monitor criteria configuration, the apparatus may collect the inference data and the monitoring data. The inference data may be used for the UE-side AI inference. In addition, the apparatus may monitor the AI model (i.e., perform the model monitoring) based on the monitor criteria configuration and the comparison result between the AI inference output and the monitoring data. Then, the apparatus may make the management decision according to the model monitoring result of the model monitoring. Then, the apparatus may transmit the decision report to the network node according to the decision result.
[0038] According to the implementations of the present disclosure, under the first proposed scheme, when the apparatus performs the model monitoring, the apparatus may compare the AI inference output and the corresponding monitoring data. In an example, the inference output may be the predicted L3 cell RSRP of the target cell at time t, and the corresponding monitoring data may be the real measurement of L3 cell RSRP of the same target cell at time t. In an example, the monitor criteria configuration may be the L3 cell RSRP difference. The apparatus may evaluate the difference between the inference output and the monitoring data. In an event that the difference is higher than a pre-configured threshold for a pre-configured time, the apparatus may generate a model monitoring result that indicates the AI model is out-of-date or unsuitable.
[0039] Based on the model monitoring result, the UE may make the management decision. The management decision may comprise at least one of an activation, a deactivation, a switching, a selection, a fallback, and keeping inference.
[0040] In an implementation, in an event that the model monitoring result shows that the current AI model is not suitable, the apparatus may make the switching decision. Based on the switching decision, a model switching request may be triggered. The model transfer or delivery procedure may be triggered. The source of the new model may be determined by the network node, a UE server, a third-party server, and / or an over-the-top (OTT) server.
[0041] In another implementation, in an event that the model monitoring result shows that the current AI model is not suitable, the apparatus may make the model selection decision. Based on the selection decision, the apparatus may reselect the AI model that is pre-downloaded and saved on the apparatus side (e.g., UE side) . Then, the new model may be applied.
[0042] In another implementation, in an event that the model monitoring result shows that the current AI model is not suitable, the apparatus may make a fallback decision. Based on the fallback decision, the apparatus may deal with the original task / or function by legacy non-AI approaches. In an example, the task may be the UE mobility, and the legacy non-AI approaches may be the L3 HO, CHO, LTM, etc.
[0043] After the management decision is made, the apparatus may transmit the decision report to the network node. The management decision may trigger additional NW action. In an implementation, the management decision may be used to ask the apparatus to fall back to non-AI approaches. The original task may use the AI prediction to reduce the measurement overhead. With the help of AI prediction, some of measurement can be skipped. After receiving the fallback decision, the network node may reconfigure the measurement configuration to the apparatus such that the apparatus performs measurement without any measurement reduction.
[0044] FIG. 2 illustrates an example scenario 200 for an AI model management process under a first proposed scheme for AI model management in accordance with implementations of the present disclosure. Scenario 200 involves an apparatus (e.g., UE) and a network node (e.g., (macro / micro) base station, TRP) of a wireless network (e.g., an LTE network, a 5G / NR network, an IoT network, or a 6G network) . Referring to FIG. 2, the network node may transmit a configuration to the UE. Then, the apparatus may perform a measurement according to the configuration to generate inference data and monitoring data. Then, the apparatus may perform the AI inference (i.e., the inference of FIG. 2) according to the inference data to generate an inference result. Then, the apparatus may perform the model monitoring (i.e., the monitor of FIG. 2) according to the inference result and the monitoring data to generate a model monitoring result. Then, the apparatus may determine the decision result (i.e., the decision of FIG. 2) according to the model monitoring result. Then, the apparatus may transmit a decision report to the network node according to the decision result.
[0045] Under a second proposed scheme for power delay profile reporting, the apparatus may perform the AI inference according to the inference data (or model input or measure data) to generate an inference result. Then, the apparatus may perform the model monitoring according to the inference result and the monitoring data to generate a model monitoring result. Then, the apparatus may transmit a monitor report to the network node according to the model monitoring result. The network node may determine the decision result according to the model monitoring result. Then, the apparatus may receive an instruction from the network node. The instruction may be based on the decision result.
[0046] In the second proposed scheme, the apparatus may perform the AI inference and the model monitoring, and the network may perform the management decision. In the second proposed scheme, the configuration that the network node transmits to the apparatus may comprise the inference configuration (or the inference data configuration) , the monitor configuration (or the monitoring data configuration) , and the monitor criteria configuration. According to the configuration, the apparatus may collect inference data and monitoring data. The inference data may be used for UE-side AI inference. The apparatus may monitor the AI model and generate monitor report based on the monitor criteria configuration and the comparison of the AI inference output and the monitoring data. The apparatus may transmit the monitor report to the network node. Then, the network node may make a management decision. The network node may transmit the instruction to the network node based on the decision result.
[0047] In the second proposed scheme, the apparatus may perform model monitoring by comparing the AI inference output and the corresponding monitor data. In an example, the AI inference output may be the predicted L3 cell RSRP of the target cell at time t, and the corresponding monitoring data may be the real measurement of the L3 cell RSRP of the same target cell at time t. In addition, in an example, the monitor criteria configuration may be the L3 cell RSRP difference. The apparatus may evaluate the difference between inference output and monitoring data. In an event that the difference remains higher than a pre-configured threshold for a pre-configured time, the apparatus may generate a monitor report that indicates the AI model is out-of-date or unsuitable.
[0048] In the second proposed scheme, the apparatus may transmit the monitor report to the network node. The network node may make the management decision (e.g., at least one of activation, deactivation, switching, selection, fallback, and keeping inference) . The network node may transmit the decision result of the management decision to the apparatus to trigger the corresponding operations or actions.
[0049] In an implementation, in an event that the model monitoring result shows that the current AI model is not suitable, the network node may make the switching decision. Based on the switching decision, a model switching request may be triggered. The model transfer or delivery procedure may be triggered. The source of the new model may be determined by the network node, a UE server, a third-party server, and / or an OTT server.
[0050] In another implementation, in an event that the model monitoring result shows that the current AI model is not suitable, the network node may make the model selection decision and transmit the decision result to the apparatus. The apparatus may reselect a new AI model that is pre-downloaded and saved on the UE side. Then, the new model will be applied.
[0051] In another embodiment, in an event that the model monitoring result shows that the current AI model is not suitable, the network node may make the fallback decision and transmit the decision result to the apparatus. The apparatus may deal with the original task or function by legacy non-AI approaches. In an example, the task may be the UE mobility, and the legacy non-AI approaches may be the L3 HO, CHO, LTM, etc.
[0052] FIG. 3 illustrates an example scenario 300 for an AI model management process under a second proposed scheme for AI model management in accordance with implementations of the present disclosure. Scenario 300 involves an apparatus (e.g., UE) and a network node (e.g., (macro / micro) base station, TRP) of a wireless network (e.g., an LTE network, a 5G / NR network, an IoT network, or a 6G network) . Referring to FIG. 3, the network node may transmit a configuration to the UE. Then, the apparatus may perform a measurement according to the configuration to generate inference data and monitoring data. Then, the apparatus may perform the AI inference (i.e., the inference of FIG. 3) according to the inference data to generate an inference result. Then, the apparatus may perform the model monitoring (i.e., the monitor of FIG. 3) according to the inference result and the monitoring data to generate a model monitoring result. Then, the apparatus may transmit a monitor report to the network node according to the model monitoring result. The network node may determine the decision result (i.e., the decision of FIG. 3) according to the model monitoring result. Then, the apparatus may receive an instruction from the network node. The instruction may be based on the decision result.
[0053] Under a third proposed scheme for power delay profile reporting, the apparatus may perform the AI inference according to the inference data (or model input or measure data) to generate an inference result. Then, the apparatus may transmit a measurement report to the network node according to the inference result. The network node may perform the model monitoring according to the inference result and the monitoring data to generate a model monitoring result. In addition, the network node may determine the decision result according to the model monitoring result. Then, the apparatus may receive an instruction from the network node. The instruction may be based on the decision result.
[0054] In the third proposed scheme, the apparatus may perform the AI inference, and the network node may perform the model monitoring and the management decision. In the third proposed scheme, the configuration that the network node transmits to the apparatus may comprise the inference configuration (or the inference data configuration) and the monitor configuration (or the monitoring data configuration) . According to the configuration, the apparatus may collect the inference data and the monitoring data. The inference data may be used for UE-side AI inference. The apparatus may transmit the AI inference output and monitoring data to the network node. The network node may monitor the AI model by comparing the AI inference output and the monitoring data. Then, the network node may make the management decision based on the model monitoring result and transmit the instruction to the apparatus based on the decision result.
[0055] In the third proposed scheme, the network node may perform model monitoring by comparing the AI inference output and the corresponding monitor data. In an example, the inference output may be the predicted L3 cell RSRP of the target cell at time t, and the corresponding monitoring data could be the real measurement of L3 cell RSRP of the same target cell at time t. In an example, the monitor criteria configuration may be the L3 cell RSRP difference. The network node may evaluate the difference between the AI inference output and the monitoring data. In an event that the difference remains higher than a pre-configured threshold for a pre-configured time, the network node may generate a monitoring result that indicates the AI model is out-of-date or unsuitable.
[0056] FIG. 4 illustrates an example scenario 400 for an AI model management process under a third proposed scheme for AI model management in accordance with implementations of the present disclosure. Scenario 400 involves an apparatus (e.g., UE) and a network node (e.g., (macro / micro) base station, TRP) of a wireless network (e.g., an LTE network, a 5G / NR network, an IoT network, or a 6G network) . Referring to FIG. 4, the network node may transmit a configuration to the UE. Then, the apparatus may perform a measurement according to the configuration to generate inference data and monitoring data. Then, the apparatus may perform the AI inference (i.e., the inference of FIG. 4) according to the inference data to generate an inference result. Then, the apparatus may transmit a measurement report to the network node according to the inference result. The network node may perform the model monitoring (i.e., the monitor of FIG. 4) according to the inference result and the monitoring data to generate a model monitoring result. In addition, the network node may determine the decision result (i.e., the decision of FIG. 4) according to the model monitoring result. Then, the apparatus may receive an instruction from the network node. The instruction may be based on the decision result.
[0057] Under a fourth proposed scheme for power delay profile reporting, the apparatus may transmit a measurement report to the network node according to the inference data (or model input or measure data) and the monitoring data. The network node may perform the AI inference according to the inference data to generate an inference result. In addition, the network node may perform model monitoring according to the inference result and the monitoring data to generate a model monitoring result. In addition, the network node may determine the decision result according to the model monitoring result. Then, the apparatus may receive an instruction from the network node. The instruction may be based on the decision result.
[0058] In the fourth proposed scheme, the network node may perform the AI inference, the model monitoring, and the management decision. In the fourth proposed scheme, the configuration that the network node transmits to the apparatus may comprise the inference configuration (or the inference data configuration) and the monitor configuration (or the monitoring data configuration) . According to the configuration, the apparatus may collect the inference data and the monitoring data, and transmit the measurement report to the network node according to the inference data and the monitoring data. The inference data may be used for NW-side AI inference. The network node may monitor the AI model by comparing the AI inference output and the monitoring data. Then, the network node may make the management decision based on the model monitoring result and transmit the instruction to the apparatus based on the decision result.
[0059] In the fourth proposed scheme, the AI inference, the model monitoring, and the management decision may be performed on the NW side. The apparatus may be involved only in collecting the required data and transmitting the required data to the network node. The algorithm of model monitoring and management decision may be transparent to the apparatus.
[0060] FIG. 5 illustrates an example scenario 500 for an AI model management process under a fourth proposed scheme for AI model management in accordance with implementations of the present disclosure. Scenario 500 involves an apparatus (e.g., UE) and a network node (e.g., (macro / micro) base station, TRP) of a wireless network (e.g., an LTE network, a 5G / NR network, an IoT network, or a 6G network) . Referring to FIG. 5, the network node may transmit a configuration to the UE. Then, the apparatus may perform a measurement according to the configuration to generate inference data and monitoring data. Then, the apparatus may transmit a measurement report to the network node according to the inference data and the monitoring data. The network node may perform the AI inference (i.e., the inference of FIG. 5) according to the inference data to generate an inference result. In addition, the network node may perform model monitoring (i.e., the monitor of FIG. 5) according to the inference result and the monitoring data to generate a model monitoring result. In addition, the network node may determine the decision result (i.e., the decision of FIG. 5) according to the model monitoring result. Then, the apparatus may receive an instruction from the network node. The instruction may be based on the decision result.
[0061] The AI model and the AI model management procedures provided in the present disclosure may be applied to different use cases. FIG. 6 illustrates an example scenario 600 for an inference data configuration (or inference configuration) and a monitor data configuration (or monitor configuration) in accordance with implementations of the present disclosure. Scenario 600 involves an apparatus (e.g., UE) and a network node (e.g., (macro / micro) base station, TRP) of a wireless network (e.g., an LTE network, a 5G / NR network, an IoT network, or a 6G network) . Referring to FIG. 6, the measurement overhead can be reduced based on the corresponding inference configuration and monitoring configuration. The inference data configuration (or inference configuration) may indicate observation information and / or predicted information. The observation information may comprise the period and offset of a target MO, e.g., SSB management timing configuration (SMTC) information, and the predicted information may comprise the period and offset of the target MO. Based on the configuration, the apparatus may measure the RSRP of the observed MO as the AI model input and predict the RSRP of the predicted MO. In addition, the monitoring data configuration (or monitoring configuration) may indicate the information of ground-truth measurement. The information of ground-truth measurement may comprise the period and offset of a target MO. The apparatus may perform model monitoring by comparing the difference between the AI inference output (e.g., predicted RSRP of target MO) and the monitoring ground-truth. In an implementation, the monitor criteria configuration may be a given threshold. In an event that the monitoring result (e.g., the RSRP difference between predicted RSRP and monitoring ground-truth RSRP) is higher than the threshold, it may imply that the AI model is out-of-date or not applicable.
[0062] According to an implementation of the present disclosure, the model monitoring procedure may be applied for the AI inter-frequency prediction use case. FIG. 7 illustrates an example scenario 700 for an inter-frequency prediction in accordance with implementations of the present disclosure. Scenario 700 involves an apparatus (e.g., UE) and a network node (e.g., (macro / micro) base station, TRP) of a wireless network (e.g., an LTE network, a 5G / NR network, an IoT network, or a 6G network) . Referring to FIG. 7, in Step 1, the apparatus may transmit the UE capability report to the network node to report its UE capability. For example, the apparatus may report that the apparatus supports the inter-frequency prediction from frequency F1 to frequency F2 and / or from frequency F1 to frequency F3. Then, in Step 2, the network node may provide the inference configuration to the apparatus to indicate the information about the inter-frequency prediction. For example, the inference configuration may comprise the information of the measure object in F1 and the prediction object in F2. In another example, the apparatus may only report the capability of inter-frequency prediction. The network node may also only configure which frequency the apparatus should predict. That is, the network node may not explicitly indicate which frequency should be measured and used for AI inference. Then, in Step 3, the apparatus may report applicability based on the inference configuration, e.g., via RRCReconfigurationComplete message. In Step 4, if the apparatus determines that it can support or has an AI model that can be activated for this inference configuration, the apparatus may transmit the report indicating the applicability to the network node and activate the AI inference. In addition, the network node may configure the corresponding monitoring configuration for the apparatus. Therefore, in Step 5, the apparatus can monitor the prediction accuracy, e.g., the RSRP difference of the predicted object. For example, the apparatus may monitor the ground-truth of the predicted measurement object in F2 by comparing the AI inference output and the ground-truth to check whether the AI model works well on the prediction of the measurement object in frequency F2 or not. In Step 6, in an event the accuracy degrades to a threshold, the apparatus may decide to deactivate the inference operation, which is informed to the network node. For example, in an event that the RSRP difference between prediction and ground-truth is larger than a given threshold (which can be configured by the network node) , the apparatus may deactivate the AI model and report to the network node. In an implementation, the apparatus may go back to the non-AI approach, which means measuring the measurement object in frequency F2 instead of prediction based on frequency F1. In an implementation, the network node may reduce the reference signal (e.g., SSB) transmission on F2 if the apparatus can perform good prediction based on its AI model. In an implementation, there should be no RS at all at the prediction objective. The RS may need to be transmitted in a sparse way. Illustrative Implementations
[0063] FIG. 8 illustrates an example communication system 800 having at least an example communication apparatus 810 and an example network apparatus 820 in accordance with an implementation of the present disclosure. Each of communication apparatus 810 and network apparatus 820 may perform various functions to implement schemes, techniques, processes and methods described herein pertaining to AI model management, including the various schemes described above with respect to various proposed designs, concepts, schemes and methods described above and with respect to user equipment and network apparatus in mobile communications, including scenarios / schemes described above as well as process 900 and process 1000 described below.
[0064] Communication apparatus 810 may be a part of an electronic apparatus, which may be a UE such as a portable or mobile apparatus, a wearable apparatus, a wireless communication apparatus or a computing apparatus. For instance, communication apparatus 810 may be implemented in a smartphone, a smartwatch, a personal digital assistant, an electronic control unit (ECU) in a vehicle, a digital camera, or a computing equipment such as a tablet computer, a laptop computer or a notebook computer. Communication apparatus 810 may also be a part of a machine type apparatus, which may be an IoT, NB-IoT, eMTC, IIoT UE such as an immobile or a stationary apparatus, a home apparatus, a roadside unit (RSU) , a wire communication apparatus or a computing apparatus. For instance, communication apparatus 810 may be implemented in a smart thermostat, a smart fridge, a smart door lock, a wireless speaker or a home control center. Alternatively, communication apparatus 810 may be implemented in the form of one or more integrated-circuit (IC) chips such as, for example and without limitation, one or more single-core processors, one or more multi-core processors, one or more reduced-instruction set computing (RISC) processors, or one or more complex-instruction-set-computing (CISC) processors. Communication apparatus 810 may include at least some of those components shown in FIG. 8 such as a processor 812, for example. Communication apparatus 810 may further include one or more other components not pertinent to the proposed schemes of the present disclosure (e.g., internal power supply, display device and / or user interface device) , and, thus, such component (s) of communication apparatus 810 are neither shown in FIG. 8 nor described below in the interest of simplicity and brevity.
[0065] Network apparatus 820 may be a part of an electronic apparatus, which may be a network node such as a satellite, a BS, a small cell, a router or a gateway of an IoT network. For instance, network apparatus 820 may be implemented in a satellite or an eNB / gNB / TRP in a 4G / 5G / B5G / 6G, NR, IoT, NB-IoT or IIoT network. Alternatively, network apparatus 820 may be implemented in the form of one or more IC chips such as, for example and without limitation, one or more single-core processors, one or more multi-core processors, or one or more RISC or CISC processors. Network apparatus 820 may include at least some of those components shown in FIG. 8 such as a processor 822, for example. Network apparatus 820 may further include one or more other components not pertinent to the proposed scheme of the present disclosure (e.g., internal power supply, display device and / or user interface device) , and, thus, such component (s) of network apparatus 820 are neither shown in FIG. 8 nor described below in the interest of simplicity and brevity.
[0066] In one aspect, each of processor 812 and processor 822 may be implemented in the form of one or more single-core processors, one or more multi-core processors, or one or more CISC processors. That is, even though a singular term “a processor” is used herein to refer to processor 812 and processor 822, each of processor 812 and processor 822 may include multiple processors in some implementations and a single processor in other implementations in accordance with the present disclosure. In another aspect, each of processor 812 and processor 822 may be implemented in the form of hardware (and, optionally, firmware) with electronic components including, for example and without limitation, one or more transistors, one or more diodes, one or more capacitors, one or more resistors, one or more inductors, one or more memristors and / or one or more varactors that are configured and arranged to achieve specific purposes in accordance with the present disclosure. In other words, in at least some implementations, each of processor 812 and processor 822 is a special-purpose machine specifically designed, arranged and configured to perform specific tasks, including AI model management, in a device (e.g., as represented by communication apparatus 810) and a network node (e.g., as represented by network apparatus 820) in accordance with various implementations of the present disclosure.
[0067] In some implementations, communication apparatus 810 may also include a transceiver 816 coupled to processor 812 and capable of wirelessly transmitting and receiving data. In some implementations, transceiver 816 may be capable of wirelessly communicating with different types of UEs and / or wireless networks of different radio access technologies (RATs) . In some implementations, transceiver 816 may comprise a main receiver and a low-power receiver. In some implementations, transceiver 816 may be equipped with a plurality of antenna ports (not shown) such as, for example, four antenna ports. That is, transceiver 816 may be equipped with multiple transmit antennas and multiple receive antennas for multiple-input multiple-output (MIMO) wireless communications. In some implementations, network apparatus 820 may also include a transceiver 826 coupled to processor 822. Transceiver 826 may include a transceiver capable of wirelessly transmitting and receiving data. In some implementations, transceiver 826 may be capable of wirelessly communicating with different types of UEs of different RATs. In some implementations, transceiver 826 may be equipped with a plurality of antenna ports (not shown) such as, for example, four antenna ports. That is, transceiver 826 may be equipped with multiple transmit antennas and multiple receive antennas for MIMO wireless communications.
[0068] In some implementations, communication apparatus 810 may further include a memory 814 coupled to processor 812 and capable of being accessed by processor 812 and storing data therein. In some implementations, network apparatus 820 may further include a memory 824 coupled to processor 822 and capable of being accessed by processor 822 and storing data therein. Each of memory 814 and memory 824 may include a type of random-access memory (RAM) such as dynamic RAM (DRAM) , static RAM (SRAM) , thyristor RAM (T-RAM) and / or zero-capacitor RAM (Z-RAM) . Alternatively, or additionally, each of memory 814 and memory 824 may include a type of read-only memory (ROM) such as mask ROM, programmable ROM (PROM) , erasable programmable ROM (EPROM) and / or electrically erasable programmable ROM (EEPROM) . Alternatively, or additionally, each of memory 814 and memory 824 may include a type of non-volatile random-access memory (NVRAM) such as flash memory, solid-state memory, ferroelectric RAM (FeRAM) , magnetoresistive RAM (MRAM) and / or phase-change memory.
[0069] Each of communication apparatus 810 and network apparatus 820 may be a communication entity capable of communicating with each other using various proposed schemes in accordance with the present disclosure. For illustrative purposes and without limitation, descriptions of capabilities of communication apparatus 810, as a UE, and network apparatus 820, as a network node (e.g., TRP) , are provided below with process 900 and process 1000. Illustrative Processes
[0070] FIG. 9 illustrates an example process 900 in accordance with an implementation of the present disclosure. Process 900 may be an example implementation of above scenarios / schemes, whether partially or completely, with respect to AI model management with the present disclosure. Process 900 may represent an aspect of implementation of features of communication apparatus 810. Process 900 may include one or more operations, actions, or functions as illustrated by one or more of blocks 910, 920, and 930. Although illustrated as discrete blocks, various blocks of process 900 may be divided into additional blocks, combined into fewer blocks, or eliminated, depending on the desired implementation. Moreover, the blocks of process 900 may be executed in the order shown in FIG. 9 or, alternatively, in a different order. Solely for illustrative purposes and without limitation, process 900 is described below in the context of communication apparatus 810. Process 900 may begin at block 910.
[0071] At block 910, process 900 may involve processor 812 of communication apparatus 810 receiving, via transceiver 816, a configuration from a network node. Process 900 may proceed from block 910 to block 920.
[0072] At block 920, process 900 may involve processor 812 of communication apparatus 810 performing a measurement according to the configuration to generate inference data and monitoring data for an AI inference and a model monitoring. Process 900 may proceed from block 920 to block 930.
[0073] At block 930, process 900 may involve processor 812 of communication apparatus 810 determining an operation for AI model management according to a decision result for the model monitoring.
[0074] In some implementations, the configuration may comprise at least one of an inference configuration, a monitor configuration, and a monitor criteria configuration.
[0075] In some implementations, the inference configuration may comprise at least one of information of an observed target and information of a predicted target.
[0076] In some implementations, the monitor configuration may comprise information of a ground-truth target.
[0077] In some implementations, the monitor criteria configuration may comprise an L1 or L3 cell RSRP difference.
[0078] In some implementations, process 900 may involve processor 812 transmitting, via transceiver 816, a measurement report to the network node according to the inference data and the monitoring data. Process 900 may involve processor 812 receiving, via transceiver 816, an instruction from the network node, wherein the instruction is based on the decision result.
[0079] In some implementations, process 900 may involve processor 812 performing the AI inference according to the inference data to generate an inference result. Process 900 may involve processor 812 performing the model monitoring according to the inference result and the monitoring data to generate a model monitoring result. Process 900 may involve processor 812 determining the decision result according to the model monitoring result. Process 900 may involve processor 812 transmitting, via transceiver 816, a decision report to the network node according to the decision result.
[0080] In some implementations, process 900 may involve processor 812 performing the AI inference according to the inference data to generate an inference result. Process 900 may involve processor 812 performing the model monitoring according to the inference result and the monitoring data to generate a model monitoring result. Process 900 may involve processor 812 transmitting, via transceiver 816, a monitor report to the network node according to the model monitoring result. Process 900 may involve processor 812 receiving, via transceiver 816, an instruction from the network node. The instruction may be based on the decision result.
[0081] In some implementations, process 900 may involve processor 812 performing the AI inference according to the inference data to generate an inference result. Process 900 may involve processor 812 transmitting, via transceiver 816, a measurement report to the network node according to the inference result. Process 900 may involve processor 812 receiving, via transceiver 816, an instruction from the network node. The instruction may be based on the decision result.
[0082] In some implementations, the operation comprises at least one of a fallback operation, a deactivate operation, and a switch operation.
[0083] FIG. 10 illustrates an example process 1000 in accordance with another implementation of the present disclosure. Process 1000 may be an example implementation of above scenarios / schemes, whether partially or completely, with respect to AI model management with the present disclosure. Process 1000 may represent an aspect of implementation of features of network apparatus 820. Process 1000 may include one or more operations, actions, or functions as illustrated by one or more of blocks 1010 and 1020. Although illustrated as discrete blocks, various blocks of process 1000 may be divided into additional blocks, combined into fewer blocks, or eliminated, depending on the desired implementation. Moreover, the blocks of process 1000 may be executed in the order shown in FIG. 10 or, alternatively, in a different order. Solely for illustrative purposes and without limitation, process 1000 is described below in the context of network apparatus 820. Process 1000 may begin at block 1010.
[0084] At block 1010, process 1000 may involve processor 822 of network apparatus 820 transmitting, via transceiver 826, a configuration for collecting inference data and monitoring data for an AI inference and a model monitoring to a UE. Process 1000 may proceed from block 1010 to block 1020.
[0085] At block 1020, process 1000 may involve processor 822 of network apparatus 820 determining an operation for AI model management according to a decision result for the model monitoring.
[0086] In some implementations, the configuration may comprise at least one of an inference configuration, a monitor configuration, and a monitor criteria configuration.
[0087] In some implementations, the inference configuration may comprise at least one of information of an observed target and information of a predicted target.
[0088] In some implementations, the monitor configuration may comprise information of a ground-truth target.
[0089] In some implementations, the monitor criteria configuration may comprise an L1 or L3 cell RSRP difference.
[0090] In some implementations, process 1000 may involve processor 822 receiving, via transceiver 826, a measurement report from the UE. The measurement report may comprise the inference data and the monitoring data. Process 1000 may involve processor 822 performing the AI inference according to the inference data to generate an inference result. Process 1000 may involve processor 822 performing the model monitoring according to the inference result and the monitoring data to generate a model monitoring result. Process 1000 may involve processor 822 determining the decision result according to the model monitoring result. Process 1000 may involve processor 822 transmitting, via transceiver 826, an instruction to the UE according to the decision result.
[0091] In some implementations, process 1000 may involve processor 822 receiving, via transceiver 826, a decision report from the UE. The decision report may comprise the decision result.
[0092] In some implementations, process 1000 may involve processor 822 receiving, via transceiver 826, a monitor report from the UE, wherein the monitor report comprises a model monitoring result. Process 1000 may involve processor 822 determining the decision result according to the model monitoring result. Process 1000 may involve processor 822 transmitting, via transceiver 826, an instruction to the UE according to the decision result.
[0093] In some implementations, process 1000 may involve processor 822 receiving, via transceiver 826, a measurement report from the UE. The measurement report may comprise an inference result and the monitoring data. Process 1000 may involve processor 822 performing the model monitoring according to the inference result and the monitoring data to generate a model monitoring result. Process 1000 may involve processor 822 determining the decision result according to the model monitoring result. Process 1000 may involve processor 822 transmitting, via transceiver 826, an instruction to the UE according to the decision result.
[0094] In some implementations, the operation may comprise at least one of a fallback operation, a deactivate operation, and a switch operation. Additional Notes
[0095] The herein-described subject matter sometimes illustrates different components contained within, or connected with, different other components. It is to be understood that such depicted architectures are merely examples, and that in fact many other architectures can be implemented which achieve the same functionality. In a conceptual sense, any arrangement of components to achieve the same functionality is effectively "associated" such that the desired functionality is achieved. Hence, any two components herein combined to achieve a particular functionality can be seen as "associated with" each other such that the desired functionality is achieved, irrespective of architectures or intermedial components. Likewise, any two components so associated can also be viewed as being "operably connected" , or "operably coupled" , to each other to achieve the desired functionality, and any two components capable of being so associated can also be viewed as being "operably couplable" , to each other to achieve the desired functionality. Specific examples of operably couplable include but are not limited to physically mateable and / or physically interacting components and / or wirelessly interactable and / or wirelessly interacting components and / or logically interacting and / or logically interactable components.
[0096] Further, with respect to the use of substantially any plural and / or singular terms herein, those having skill in the art can translate from the plural to the singular and / or from the singular to the plural as is appropriate to the context and / or application. The various singular / plural permutations may be expressly set forth herein for sake of clarity.
[0097] Moreover, it will be understood by those skilled in the art that, in general, terms used herein, and especially in the appended claims, e.g., bodies of the appended claims, are generally intended as “open” terms, e.g., the term “including” should be interpreted as “including but not limited to, ” the term “having” should be interpreted as “having at least, ” the term “includes” should be interpreted as “includes but is not limited to, ” etc. It will be further understood by those within the art that if a specific number of an introduced claim recitation is intended, such an intent will be explicitly recited in the claim, and in the absence of such recitation no such intent is present. For example, as an aid to understanding, the following appended claims may contain usage of the introductory phrases "at least one" and "one or more" to introduce claim recitations. However, the use of such phrases should not be construed to imply that the introduction of a claim recitation by the indefinite articles "a" or "an" limits any particular claim containing such introduced claim recitation to implementations containing only one such recitation, even when the same claim includes the introductory phrases "one or more" or "at least one" and indefinite articles such as "a" or "an, " e.g., “a” and / or “an” should be interpreted to mean “at least one” or “one or more; ” the same holds true for the use of definite articles used to introduce claim recitations. In addition, even if a specific number of an introduced claim recitation is explicitly recited, those skilled in the art will recognize that such recitation should be interpreted to mean at least the recited number, e.g., the bare recitation of "two recitations, " without other modifiers, means at least two recitations, or two or more recitations. Furthermore, in those instances where a convention analogous to “at least one of A, B, and C, etc. ” is used, in general such a construction is intended in the sense one having skill in the art would understand the convention, e.g., “a system having at least one of A, B, and C” would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together, etc. In those instances where a convention analogous to “at least one of A, B, or C, etc. ” is used, in general such a construction is intended in the sense one having skill in the art would understand the convention, e.g., “a system having at least one of A, B, or C” would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together, etc. It will be further understood by those within the art that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the description, claims, or drawings, should be understood to contemplate the possibilities of including one of the terms, either of the terms, or both terms. For example, the phrase “A or B” will be understood to include the possibilities of “A” or “B” or “A and B. ”
[0098] From the foregoing, it will be appreciated that various implementations of the present disclosure have been described herein for purposes of illustration, and that various modifications may be made without departing from the scope and spirit of the present disclosure. Accordingly, the various implementations disclosed herein are not intended to be limiting, with the true scope and spirit being indicated by the following claims.
Claims
1.A method, comprising:receiving, by a processor of an apparatus, a configuration from a network node;performing, by the processor, a measurement according to the configuration to generate inference data and monitoring data for an artificial intelligence (AI) inference and a model monitoring; anddetermining, by the processor, an operation for AI model management according to a decision result for the model monitoring.2.The method of Claim 1, wherein the configuration comprises at least one of an inference configuration, a monitor configuration, and a monitor criteria configuration.3.The method of Claim 2, wherein the inference configuration comprises at least one of information of an observed target and information of a predicted target.4.The method of Claim 2, wherein the monitor configuration comprises information of a ground-truth target.5.The method of Claim 2, wherein the monitor criteria configuration comprises a layer 1 (L1) or layer 3 (L3) cell reference signal received power (RSRP) difference.6.The method of Claim 1, further comprising:transmitting, by the processor, a measurement report to the network node according to the inference data and the monitoring data; andreceiving, by the processor, an instruction from the network node, wherein the instruction is based on the decision result.7.The method of Claim 1, further comprising:performing, by the processor, the AI inference according to the inference data to generate an inference result;performing, by the processor, the model monitoring according to the inference result and the monitoring data to generate a model monitoring result; anddetermining, by the processor, the decision result according to the model monitoring result; andtransmitting, by the processor, a decision report to the network node according to the decision result.8.The method of Claim 1, further comprising:performing, by the processor, the AI inference according to the inference data to generate an inference result;performing, by the processor, the model monitoring according to the inference result and the monitoring data to generate a model monitoring result; andtransmitting, by the processor, a monitor report to the network node according to the model monitoring result; andreceiving, by the processor, an instruction from the network node, wherein the instruction is based on the decision result.9.The method of Claim 1, further comprising:performing, by the processor, the AI inference according to the inference data to generate an inference result;transmitting, by the processor, a measurement report to the network node according to the inference result; andreceiving, by the processor, an instruction from the network node, wherein the instruction is based on the decision result.10.The method of Claim 1, wherein the operation comprises at least one of a fallback operation, a deactivate operation, and a switch operation.11.A method, comprising:transmitting, by a processor of a network node, a configuration for collecting inference data and monitoring data for an artificial intelligence (AI) inference and a model monitoring to a user equipment (UE) ; anddetermining, by the processor, an operation for AI model management according to a decision result for the model monitoring.12.The method of Claim 11, wherein the configuration comprises at least one of an inference configuration, a monitor configuration, and a monitor criteria configuration.13.The method of Claim 12, wherein the inference configuration comprises at least one of information of an observed target and information of a predicted target.14.The method of Claim 12, wherein the monitor configuration comprises information of a ground-truth target.15.The method of Claim 12, wherein the monitor criteria configuration comprises a layer 1 (L1) or layer 3 (L3) cell reference signal received power (RSRP) difference.16.The method of Claim 11, further comprising:receiving, by the processor, a measurement report from the UE, wherein the measurement report comprises the inference data and the monitoring data;performing, by the processor, the AI inference according to the inference data to generate an inference result;performing, by the processor, the model monitoring according to the inference result and the monitoring data to generate a model monitoring result;determining, by the processor, the decision result according to the model monitoring result; andtransmitting, by the processor, an instruction to the UE according to the decision result.17.The method of Claim 11, further comprising:receiving, by the processor, a decision report from the UE, wherein the decision report comprises the decision result.18.The method of Claim 11, further comprising:receiving, by the processor, a monitor report from the UE, wherein the monitor report comprises a model monitoring result;determining, by the processor, the decision result according to the model monitoring result; andtransmitting, by the processor, an instruction to the UE according to the decision result.19.The method of Claim 11, further comprising:receiving, by the processor, a measurement report from the UE, wherein the measurement report comprises an inference result and the monitoring data;performing, by the processor, the model monitoring according to the inference result and the monitoring data to generate a model monitoring result;determining, by the processor, the decision result according to the model monitoring result; andtransmitting, by the processor, an instruction to the UE according to the decision result.20.The method of Claim 11, wherein the operation comprises at least one of a fallback operation, a deactivate operation, and a switch operation.