Methods and apparatus of ai / ML based RRM measurement for wireless communication systems
AI/ML-based RRM measurement prediction methods address the inefficiencies in 5G NR networks by optimizing UE mobility through reduced measurement overhead and power consumption, ensuring seamless handovers and improved network performance.
Patent Information
- Application Number
- PCT/CN2023/142377
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-12-27
- Publication Date
- 2025-07-03
AI Technical Summary
Conventional wireless communication systems face increased measurement overhead and power consumption due to enhanced UE mobility, particularly in 5G NR networks, which is exacerbated by frequent handovers and higher cell densities, leading to inefficiencies in signaling and latency.
Employing AI/ML-based RRM measurement prediction methods that utilize beam sweeping and AI/ML models to forecast beam quality, reducing unnecessary measurements and optimizing reporting criteria for UE mobility, including L3 HO and LTM mechanisms.
Reduces measurement overhead and power consumption while maintaining service continuity and quality by providing proactive AI/ML-driven mobility management, enhancing network performance and user experience.
Smart Images

Figure CN2023142377_03072025_PF_FP_ABST
Abstract
Description
METHODS AND APPARATUS OF AI / ML BASED RRM MEASUREMENT FOR WIRELESS COMMUNICATION SYSTEMSFIELD
[0001] The present disclosure relates generally to communication systems, and more particularly, the method of AI / ML based RRM measurement prediction for wireless communication systems.BACKGROUND
[0002] In the conventional network of the 3rd generation partnership project (3GPP) 5G new radio (NR) , considered UE mobility, UE keeps measures and reports radio resource signal environment quality to maintain service continuity and quality. Legacy HO design controlled by a series of L3 procedures including RRM measurement and RRC Reconfiguration, which involves amount of signaling and latency. Conditional handover is introduced to reduces signaling overhead by allowing the UE to make a handover decision based on predefined conditions, leading to a faster HO. Dual Active Protocol Stack (DAPS) aims to minimize interruption by maintaining two active protocol stacks, allowing the UE to communicate with both the source and target cells simultaneously during the handover process. To further reduce the latency, overhead, and interruption time during UE mobility, the mobility mechanism L1 / L2 based inter-cell mobility (LTM, L1L2-triggered Mobility) is enhanced to enable a serving cell to change via beam management with L1 / L2 signaling.
[0003] Artificial Intelligence (AI) and Machine Learning (ML) are increasingly being integrated into wireless networks to enhance performance, elevate user experience, and diminish complexity and overhead. In Release 18 (Rel-18) , AI / ML technologies have been harnessed to forecast the top-K beams, either in the time or spatial domain, which aids in curtailing the measurement overhead. As mobility increases, that is, the UE moves faster / more frequently or cells with higher density, the measurement overhead and power consumption will also increase accordingly. To solve the measurement overhead problem caused by enhanced mobility handover, using AI / ML predictive measurements to reduce unnecessary measurements, while providing proactive measurements for HO.
[0004] In this invention, apparatus and mechanisms are sought to use AI / ML based RRM measurement prediction for L3 HO and LTM mobility mechanisms of wireless communication systems to reduce the measurement overhead.SUMMARY
[0005] The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects, and is intended to neither identify key or critical elements of all aspects nor delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.
[0006] In an aspect of the disclosure, a method, a computer-readable medium, and an apparatus are provided. The apparatus may be a UE. The UE performs beam sweeping to measures on a first set of reference signals for first set of cells, then predicts the measurement results on a second set of reference signals for a second set of cells with one or multiple AO / ML models. The measurement results and assistance information (UE positioning / speed) can be used to construct the model input. The set of reference signals for the model label and model input are defined by spatial, temporal, frequency domain and any combinations of the domains. The AI / ML model is applied to derive the cell quality of the first set, the second set of cells or both of them, and also is applied to evaluating the reporting criteria to trigger the measurement report and sending the measurement report to the network.
[0007] To the accomplishment of the foregoing and related ends, the one or more aspects comprise the features hereinafter fully described and particularly pointed out in the claims. The following description and the annexed drawings set forth in detail certain illustrative features of the one or more aspects. These features are indicative, however, of but a few of the various ways in which the principles of various aspects may be employed, and this description is intended to include all such aspects and their equivalents.BRIEF DESCRIPTION OF THE DRAWINGS
[0008] Figure 1 illustrates a schematic system diagram illustrating an exemplary 5G new radio network in accordance with embodiments of the current invention.
[0009] Figure 2 illustrates an exemplary diagram for the A / ML based RRM radio resource of reference signal defined by spatial, temporal frequency domain and any combinations of the domains in accordance with embodiments of the current invention.
[0010] Figure 3 illustrates an exemplary overall flow of the high-level measurement model in accordance with embodiments of the current invention.
[0011] Figure 4 illustrates an exemplary overall flow for the measurement results without filtering used for AI / ML based RRM prediction in accordance with embodiments of the current invention.
[0012] Figure 5 illustrate an exemplary overall flow for the L1 measurement results (after L1 filtering) used for AI / ML based RRM prediction in accordance with embodiments of the current invention.
[0013] Figure 6 illustrate an exemplary overall flow for the L3 measurement results used for AI / ML based RRM prediction are s in accordance with embodiments of the current invention.
[0014] Figure 7 illustrate an exemplary overall flow for RRM measurements cell-quality prediction in accordance with embodiments of the current invention.
[0015] Figure 8 illustrate an exemplary overall flow for AI / ML based RRM prediction measurement procedure in accordance with embodiments of the current invention.DETAILED DESCRIPTION
[0016] The detailed description set forth below in connection with the appended drawings is intended as a description of various configurations and is not intended to represent the only configurations in which the concepts described herein may be practiced. The detailed 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 structures and components are shown in block diagram form in order to avoid obscuring such concepts.
[0017] Several aspects of telecommunication systems will now be presented with reference to various apparatus and methods. These apparatus and methods will be described in the following detailed description and illustrated in the accompanying drawings by various blocks, components, circuits, processes, algorithms, etc. (collectively referred to as “elements” ) . These elements may be implemented using electronic hardware, computer software, or any combination thereof. Whether such elements are implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system.
[0018] Aspects of the present disclosure provide methods, apparatus, processing systems, and computer readable mediums for NR (new radio access technology, or 5G technology) , 6G or other radio access technology. NR may support various wireless communication services. These services may havse different quality of service (QoS) requirements e.g. latency and reliability requirements.
[0019] Figure 1 illustrates a schematic system diagram illustrating an exemplary wireless network in accordance with embodiments of the current invention. Wireless system includes one or more fixed base infrastructure units forming a network distributed over a geographical region. The base unit may also be referred to as an access point, an access terminal, a base station, a Node- B, an eNode-B, a gNB, or by other terminology used in the art. As an example, base stations serve a number of mobile stations within a serving area, for example, a cell, or within a cell sector. In some systems, one or more base stations are coupled to a controller forming an access network that is coupled to one or more core networks. gNB 1and gNB 2 are base stations in NR, the serving area of which may or may not overlap with each other. As an example, UE1 or mobile station is only in the service area of gNB 1 and connected with gNB1. UE1 is connected with gNB1 only, gNB1 is connected with gNB 1 and 2 via Xn interface. UE2 is in the overlapping service area of gNB1 and gNB2.
[0020] Figure 1 further illustrates simplified interaction flow for UE1 and gNB2. The gNB provides measurement configuration to UE by RRC configuration signaling. The gNB can request UE to measure intra-frequency, inter-frequency measurements. UE performs measurements and based on the measurement evaluate the criteria for measurement report. The AI / ML model is used to predict the actual measurements results. In one embodiment, the AI / ML model is at UE side. UE performs measurement prediction and report the model output to gNB. gNB makes handover decision and sends handover command to UEs.
[0021] Figure 2 illustrates an exemplary diagram for the A / ML based RRM radio resource of reference signal defined by spatial, temporal frequency domain and any combinations of the domains in accordance with embodiments of the current invention. In one embodiment, UE measures the first set of reference signals for a first set of cells and the second set of reference signals for a second set of cells at training stage. In one embodiment, UE performs model training using the measurement results of the first set of reference signals for the first set of cells as model input and using the measurement results of the second set of reference signals for the second set of cells as label.
[0022] In one embodiment (as shown in sub-figure (a) ) , UE performs measurement prediction at temporal domain with one or multiple AI / ML models, which means the first set of reference signals for a first set of cells (model input) and the second set of reference signals for a second set of cells (model label) are same in frequency and spatial domain. At training stage, AI / ML based RRM model using historical measurement results [T-N, T] as model input and future measurement results [T+1, T+M] as label. UE has one or multiple trained AI / ML models, as an example, UE measures the first set of reference signals for a first set of cells at time [T-N, T] to predicts the second set of reference signals for a second set of cells at time [T, T+M] .
[0023] In one embodiment (as shown in sub-figure (b) ) , UE performs measurement prediction at spatial domain with one or multiple AI / ML models, which means the first set of reference signals for a first set of cells and the second set of reference signals for a second set of cells are same in frequency and temporal domain. The solid lines indicates the measured set of beams and the dotted line indicated the predicted set of beams. At training stage, AI / ML based RRM model using measurement results of partial beams as model input and full / the other part of beam results as label. UE has one or multiple trained AI / ML models, as an example, UE measures the first set of reference signals for a first set of cells at partial beams set with dotted line to predicts the second set of reference signals for a second set of cells of full / the other part of beam set with solid line.
[0024] In one embodiment (as shown in sub-figure (c) ) , UE performs measurement prediction at frequency domain with one or multiple AI / ML models, which means the first set of reference signals for a first set of cells and the second set of reference signals for a second set of cells are same in spatial and temporal domain. The solid line indicates the measured BWPs / frequencies, and the dotted line indicated the predicted full / other BWPs / frequencies, AI / ML based RRM model using measurement results of some BWPs / frequencies as model input and full / other BWPs / frequencies results as label. In one embodiment, the first set of reference signals and the second set of reference signals are inter-frequency. In one embodiment, the first set of reference signals and the second set of reference signals are intra-frequency.
[0025] In one embodiment, the first set of reference signals for a first set of cells and the second set of reference signals for a second set of cells are any combinations of the domains. In one embodiment, the first set of reference signals and the second set of reference signals are same in frequency domain, and UE performs measurement prediction at temporal-spatial domain with one or multiple AI / ML models which combine the sub-figure (a) and sub-figure (b) , using historical measurement results [T-N, T] of partial beams as model input and future measurement results [T+1, T+M] of the full / other part of beam results as label.
[0026] In one embodiment, the first set of reference signals for a first set of cells and the second set of reference signals for a second set of cells are same in spatial domain, and UE performs measurement prediction at temporal-frequency domain with one or multiple AI / ML models which combine combination of sub-figure (a) and sub-figure (c) , using historical measurement results [T-N, T] of some BWPs / frequencies as model input and future measurement results [T+1, T+M] of full / other BWPs / frequencies as label.
[0027] In one embodiment, the first set of reference signals for a first set of cells and the second set of reference signals for a second set of cells are same in time domain, and UE performs measurement prediction at spatial-frequency domain with one or multiple AI / ML models which combine combination of sub-figure (b) and sub-figure (c) , using measurement results of partial beams for some BWPs / frequencies as model input and the full / other part of beam results for full / other BWPs / frequencies results as label.
[0028] In one embodiment, the UE performs measurement prediction at temporal-spatial- frequency domain with one or multiple AI / ML models which combine combination of sub-figure (b) , sub-figure (b) and sub-figure (c) , using historical measurement results [T-N, T] of partial beams for some BWPs / frequencies as model input and future measurement results [T+1, T+M] of the full / other part beam results for full / other BWPs / frequencies as label.
[0029] Figure 3 illustrates an exemplary overall flow of the high-level measurement model in accordance with embodiments of the current invention. The measurement procedure as showed is an exemplary flow, including beam-level and cell-level filtering. K beams (pairs) correspond to the measurements on SSB or CSI-RS resources configured for L3 mobility by gNB and detected by UE at L1. The point A is the measurements (beam specific samples) internal to the physical layer. The Layer 1 filtering is internal layer 1 filtering of the inputs measured at point A and the implementation is up to UE. The point A1 represents the measurements reported by layer 1 to layer3 after layer 1 filtering. Beam consolidation performs consolidation of beam specific measurements to derive cell quality. Layer 3 filtering for cell quality filtering the measurements provided at point B. The point C is the measurement after processing of layer 3 filtering for cell quality. By evaluation of reporting criteria, the measurement report message D is sent on the radio interface. L3 beam filtering filter the measurement provided at point A1. Beam selection for beam reporting select X measurements from the measurements after L3 beam filtering. The behaviour of the beam consolidation / selection, Layer 3 filters, evaluation of reporting criteria, L3 beam filtering and beam selection for beam reporting is standardized and the configuration of the layer 3 filters is provided by RRC signaling.
[0030] Figure 4 illustrates an exemplary overall flow for the measurement results without filtering used for AI / ML based RRM prediction in accordance with embodiments of the current invention. In one embodiment, using measurements results of the first set of reference signa without filtering as model input, UE predicts the measurement results on the second set of reference signals first and then performs L1 and L3 filtering on the measurement results of both the first set and the second set of reference signals, performs beam consolidation / selection and finally derives the cell quality. In one embodiment, the first set of reference signal and the second set of reference signals are defined by spatial, temporal, frequency domain and any combinations of the domains.
[0031] Figure 5 illustrate an exemplary overall flow for the L1 measurement results (after L1 filtering) used for AI / ML based RRM prediction in accordance with embodiments of the current invention. In one embodiment, using measurements results of the first set of reference signal after L1 filtering as model input, UE predicts the measurement results on the second set of reference signals first and then performs L3 filtering on the measurement results of both the first set and the second set of reference signals, performs beam consolidation / selection and finally derives the cell quality. In one embodiment, the first set of reference signal and the second set of reference signals are defined by spatial, temporal, frequency domain and any combinations of the domains.
[0032] Figure 6 illustrate an exemplary overall flow for the L3 measurement results used for AI / ML based RRM prediction are s in accordance with embodiments of the current invention. In one embodiment, using measurements results of the first set of reference signal after L3 filtering as model input, UE predicts the measurement results on the second set of reference signals first. In one embodiment, UE performs L3 filter on the measurement results of the first set of reference signals first, performs measurement prediction on the second set of reference signals, then performs beam consolidation / selection and finally derives the cell quality. In one embodiment, the first set of reference signal and the second set of reference signals are defined by spatial, temporal, frequency domain and any combinations of the domains.
[0033] Figure 7 illustrate an exemplary overall flow for RRM measurements cell-quality prediction in accordance with embodiments of the current invention. In one embodiment, UE measures the first set of cells and predicts the measurement results of the second set of cells. In one embodiment, the UE measures the first set of cells is same with the predicted second set of cells. In one embodiment, the UE measures the first set of cells is different with the predicted second set of cells. In one embodiment (a) , the measurement results used for prediction are L1 measurement results (after L1 filtering) . In one embodiment (b) , the measurement results used for prediction are L3 measurement results (after L3 cell filtering) . In one embodiment, the first set of reference signal for the first set of cells and the second set of reference signals for the second cells are defined by spatial, temporal, frequency domain and any combinations of the domains.
[0034] Figure 8 illustrate an exemplary overall flow for AI / ML based RRM prediction measurement procedure in accordance with embodiments of the current invention. In one embodiment, for the temporal domain beam-quality prediction, the UE performs full beams sweeping at observation time window and stop beam sweeping at prediction time window. In one embodiment, for the spatial domain beam-quality prediction, the UE performs partial beam sweeping. In one embodiment, for the temporal-spatial domain beam-quality prediction, the UE performs partial beams sweeping at observation time window. In one embodiment, the UE evaluates the reporting criteria considers the measurement results which are actually measured, the measurement results which are predicted, or both of them are used, and which approach to use is configured by the network. In one embodiment, the UE reports the measurement results which are actually measured, the measurement results which are predicted, or both of them, and which approach to use is configured by the network. In one embodiment, the one or multiple AI / ML models is cell specific, i.e., one AI / ML model corresponds to one cell. In one embodiment, the one or multiple AI / ML models is cluster specific, i.e., one AI / ML model corresponds to a cluster of cells. In one embodiment, the measured RS quality can be SNR / RSRP / RSS.
[0035] It is understood that the specific order or hierarchy of blocks in the processes / flowcharts disclosed is an illustration of exemplary approaches. Based upon design preferences, it is understood that the specific order or hierarchy of blocks in the processes / flowcharts may be rearranged. Further, some blocks may be combined or omitted. The accompanying method claims present elements of the various blocks in a sample order, and are not meant to be limited to the specific order or hierarchy presented.
[0036] The previous description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects. Thus, the claims are not intended to be limited to the aspects shown herein but is to be accorded the full scope consistent with the language claims, wherein reference to an element in the singular is not intended to mean “one and only one” unless specifically so stated, but rather “one or more. ” The word “exemplary” is used herein to mean “serving as an example, instance, or illustration. ” Any aspect described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects. Unless specifically stated otherwise, the term “some” refers to one or more. Combinations such as “at least one of A, B, or C, ” “one or more of A, B, or C, ” “at least one of A, B, and C, ” “one or more of A, B, and C, ” and “A, B, C, or any combination thereof” include any combination of A, B, and / or C, and may include multiples of A, multiples of B, or multiples of C. Specifically, combinations such as “at least one of A, B, or C, ” “one or more of A, B, or C, ” “at least one of A, B, and C, ” “one or more of A, B, and C, ” and “A, B, C, or any combination thereof” may be A only, B only, C only, A and B, A and C, B and C, or A and B and C, where any such combinations may contain one or more member or members of A, B, or C. All structural and functional equivalents to the elements of the various aspects described throughout this disclosure that are known or later come to be known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed by the claims. Moreover, nothing disclosed herein is intended to be dedicated to the public regardless of whether such disclosure is explicitly recited in the claims. The words “module, ” “mechanism, ” “element, ” “device, ” and the like may not be a substitute for the word “means. ” As such, no claim element is to be construed as a means plus function unless the element is expressly recited using the phrase “means for. ”
[0037] While aspects of the present disclosure have been described in conjunction with the specific embodiments thereof that are proposed as examples, alternatives, modifications, and variations to the examples may be made. Accordingly, embodiments as set forth herein are intended to be illustrative and not limiting. There are changes that may be made without departing from the scope of the claims set forth below.
Claims
1.A method for UE to reduce RRM measurement effort and RS overhead by AI-based RRM prediction, comprising the steps of:performing measurement on a first set of reference signals for a first set of cells, wherein the radio resources of the first set of reference signals are defined by spatial, temporal, frequency domain and any combinations of the domains and the first set of the cells can be the serving cells and the neighboring cells of intra-frequency or inter-frequency;predicting the measurement results on a second set of reference signals for a second set of cells with one or multiple AI / ML models, wherein radio resources of the second set of reference signals are defined by spatial, temporal, frequency domain and any combinations of the domains and the second set of the cells can be the serving cells and the neighboring cells of intra-frequency or inter-frequency, which can be same or different from the first set of cells;deriving the cell quality of the first set, the second set of cells or both of them;evaluating the reporting criteria to trigger the measurement report and sending the measurement report to the network.2.The method of claim 1, further comprising collecting the measurement results of the first set of reference signals for the first set of cells and the second set of reference signals for the second set of cells and training one or multiple AI / ML models based on the collected measurement results with / without other assistance information.3.The method of claim 2, wherein UE performs measurement to the first set of reference signals for the first set of cells and the second set of reference signals for the second set of cells at training stage.4.The method of claim 3, wherein UE performs model training using the measurement results of the first set of reference signals for the first set of cells as model input and using the measurement results of the second set of reference signals for the second set of cells as label.5.The method of claim 1, wherein the first set of reference signals and the second set of reference signals are same in frequency and spatial domain, and UE performs measurement prediction at temporal domain with one or multiple AI / ML models, using historical measurement results [T-N, T] as model input and future measurement results [T+1, T+M] as label.6.The method of claim 1, wherein the first set of reference signals and the second set of reference signals are same in frequency and temporal domain, and UE performs measurement prediction at spatial domain with one or multiple AI / ML models, using measurement results of partial beams as model input and full / the other part of beam results as label.7.The method of claim 1, wherein the first set of reference signals and the second set of reference signals are same in spatial and temporal domain, and UE performs measurement prediction at frequency domain with one or multiple AI / ML models, using measurement results of some BWPs / frequencies as model input and full / other BWPs / frequencies results as label.8.The method of claim 1, wherein the first set of reference signals and the second set of reference signals are same in frequency domain, and UE performs measurement prediction at temporal-spatial domain with one or multiple AI / ML models, using historical measurement results [T-N, T] of partial beams as model input and future measurement results [T+1, T+M] of the full / other part of beam results as label.9.The method of claim 1, wherein the first set of reference signals and the second set of reference signals are same in spatial domain, and UE performs measurement prediction at temporal-frequency domain with one or multiple AI / ML models, using historical measurement results [T-N, T] of some BWPs / frequencies as model input and future measurement results [T+1, T+M] of full / other BWPs / frequencies as label.10.The method of claim 1, wherein the first set of reference signals and the second set of reference signals are same in time domain, and UE performs measurement prediction at spatial-frequency domain with one or multiple AI / ML models, using measurement results of partial beams for some BWPs / frequencies as model input and the full / other part of beam results for full / other BWPs / frequencies results as label.11.The method of claim 1, wherein the UE performs measurement prediction at temporal-spatial-frequency domain with one or multiple AI / ML models, using historical measurement results [T-N, T] of partial beams for some BWPs / frequencies as model input and future measurement results [T+1, T+M] of the full / other part beam results for full / other BWPs / frequencies as label.12.The method of claim 1, wherein UE measures the first set of cells and predicts the measurement results of the second set of cells.13.The method of claim 1, wherein the measurement results used for prediction are L1 measurement results (after L1 filtering) , i.e., UE predicts the measurement results on the second set of reference signals first and then performs L3 filtering on the measurement results of both the first set and the second set of reference signals.14.The method of claim 1, wherein the measurement results used for prediction are L3 measurement results, i.e., UE performs L3 filtering on the measurement results of the first set of reference signals first and then predicts the measurement results on the second set of reference signals.15.The method of claim 14, wherein UE performs L3 filter on the measurement results of the first set of reference signals first, performs measurement prediction on the second set of reference signals, performs beam consolidation / selection and finally derives the cell quality.16.The method of claim 1, wherein evaluating the reporting criteria considers the measurement results which are actually measured, the measurement results which are predicted, or both of them are used, and which approach to use is configured by the network.17.The method of claim 1, wherein measurement reporting reports the measurement results which are actually measured, the measurement results which are predicted, or both of them, and which approach to use is configured by the network.18.The method of claim 1, wherein the one or multiple AI / ML models is cell specific, i.e., one AI / ML model corresponds to one cell, or cluster specific, i.e., one AI / ML model corresponds to a cluster of cells.19.The method of claim 1, wherein the measured RS quality can be SNR / RSRP / RSS.
Citation Information
Patent Citations
Applying measurements on first frequency band to processes on second frequency band
CN115039441A
Techniques for cross-band channel prediction and reporting
CN115735339A
Terminal, wireless communication method, and base station
CN117099438A
Communication method, network device, terminal, communication system and storage medium
CN117296362A
Mobility reporting for non-serving cells based on machine learning
WO2023137684A1