Multi-stage model for ue-based ai / ml mobility

Abstract

In some embodiments, multiple artificial intelligence (AI) / machine learning (ML) models can be chained. Each model can act as a stage of the chain. The AI / ML model chain can include multiple models used by a UE. An early stage model can trigger a final stage model to run. The UE can use the AI / ML model chain to predict a measurement or mobility event and report the prediction to a network node.

Description

Technical Field

[0001] This application relates in general to wireless communication systems, including AI / ML models for mobility prediction. Background Technology

[0002] Wireless mobile communication technologies use various standards and protocols to transmit data between base stations and wireless communication devices. Wireless communication system standards and protocols may include, for example, 3GPP Long Term Evolution (LTE) (e.g., 4G), 3GPP New Radio (NR) (e.g., 5G), and the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standard for Wireless Local Area Networks (WLANs) (often referred to as Wi-Fi within the industry organization). ® ).

[0003] As envisioned by 3GPP, different wireless communication system standards and protocols can use various radio access networks (RANs) for communication between RAN base stations (sometimes referred to as RAN nodes, network nodes, or simply nodes) and wireless communication equipment called user equipment (UEs). 3GPP RANs can include, for example, Global System for Mobile Communications (GSM), Enhanced Data Rate GSM Evolution (EDGE) RAN (GERAN), Universal Terrestrial Radio Access Network (UTRAN), Evolved Universal Terrestrial Radio Access Network (E-UTRAN), and / or Next Generation Radio Access Network (NG-RAN).

[0004] Each RAN can use one or more Radio Access Technologies (RATs) to perform communication between the base station and the UE. For example, GERAN implements the GSM and / or EDGE RAT, UTRAN implements the Universal Mobile Telecommunications System (UMTS) RAT or other 3GPP RATs, E-UTRAN implements the LTE RAT (sometimes simply referred to as LTE), and NG-RAN implements the NR RAT (sometimes referred to herein as 5GRAT, 5G NR RAT, or simply NR). In some deployments, E-UTRAN may also implement the NR RAT. In some deployments, NG-RAN may also implement the LTE RAT.

[0005] The base stations used by a RAN can be associated with that RAN. An example of an E-UTRAN base station is an Evolved Universal Terrestrial Radio Access Network (E-UTRAN) Node B (also commonly referred to as Evolved Node B, Enhanced Node B, eNodeB, or eNB). An example of an NG-RAN base station is a Next Generation Node B (sometimes also called gNode B or gNB).

[0006] The RAN provides communication services to external entities through its connection with the core network (CN). For example, E-UTRAN can utilize the evolved packet core (EPC), while NG-RAN can utilize the 5G core network (5GC). Attached Figure Description

[0007] To facilitate the identification of any particular element or action in the discussion, one or more of the most significant digits in the figure reference numerals refer to the figure number in which the element was first introduced.

[0008] Figure 1 Example block diagrams illustrating two chained AI / ML models according to some implementation schemes are shown.

[0009] Figure 2 Example block diagrams of three chained AI / ML models based on some implementation schemes are shown.

[0010] Figure 3 Example block diagrams are shown for chained AI / ML models with a shared early-stage model according to some implementation schemes.

[0011] Figure 4 Example methods executed by the UE according to some implementation schemes are illustrated.

[0012] Figure 5 Example methods executed by network nodes according to some implementation schemes are illustrated.

[0013] Figure 6 An example architecture of a wireless communication system according to the implementation scheme disclosed herein is illustrated.

[0014] Figure 7 A system for performing signaling transfer between a wireless device and a network device according to an embodiment disclosed herein is illustrated. Detailed Implementation

[0015] Various implementations are described with respect to the UE. However, references to the UE are provided for illustrative purposes only. The example implementations may be used with any electronic components capable of establishing a connection to a network and configured with hardware, software, and / or firmware for exchanging information and data with the network. Therefore, the UE described herein is used to represent any suitable electronic component.

[0016] One goal of wireless networks is to improve reliability. This can include the UE performing measurements to determine the current state of its connection to network nodes. However, performing measurements comes at the cost of power consumption at the UE. Since the UE can be a battery-limited device, another goal of wireless networks is to reduce power consumption. These goals can conflict with each other. For example, performing fewer measurements may reduce power consumption, but it may also reduce the reliability of the network connection.

[0017] Artificial intelligence (AI) / machine learning (ML) can be used to reduce mobility-related power consumption while limiting the impact on network connectivity reliability. The implementations described herein use AI / ML to improve the performance of the UE or other battery-constrained devices. In some implementations, the UE can use AI / ML-based RRM measurements and / or event predictions. For example, cell-level measurement predictions may include intra-frequency and inter-frequency (UE-side and NW-side models). AI / ML can be used for inter-cell beam-level measurement predictions for L3 mobility (UE-side and NW-side models). Additionally, AI / ML can be used for handover failure / RLF prediction (UE-side model), measurement event prediction (UE-side model), and other UE-aided information for network-side models.

[0018] One objective of UE-based AI / ML measurement prediction is to reduce UE power consumption. However, while AI / ML inference may consume less energy than actual measurements, it is not without cost. This is especially true when model inference must be performed continuously. One objective of the implementation described in this paper is to reduce the computational and power requirements of UE-based AI / ML inference for mobility. While the implementation described in this paper focuses on AI / ML inference for mobility, AI / ML principles can be applied to other scenarios for battery-constrained devices.

[0019] Figure 1 Example block diagram 102 illustrates two chained AI / ML models according to some implementation schemes. This block diagram 102 illustrates a simplified operating principle of the embodiments of this disclosure. In some implementations, multiple models linked together can be used by the UE to reduce power consumption. Unlike distributed models, the UE can run both types of models. A chained model can refer to multiple models used by the UE, where a first model triggers the execution of a second model. The second model may not run unless a trigger from the first model is output. Each model in the chain can be referred to as a stage. Therefore, a chained model can also be referred to as a multi-stage model.

[0020] In the illustrated implementation, model 1 104 may output a trigger that causes model 0 106 to run. Model 0 106 may be an actual model trained to predict specific outputs (e.g., mobility events and / or measurements). For example, model 0 106 may predict mobility events such as radio link failures, handover failures, measurement events, etc. In some implementations, model 0 106 may be configured to predict measurements, so that the UE does not need to perform as many actual measurements.

[0021] Model 0 106 can be a more complex model than Model 1 104. Therefore, Model 0 106 can operate using more power than Model 1 104. In the illustrated implementation, Model 1 104 is a simplified model trained to predict when mobility-related events may occur. In some implementations, the input parameters of Model 1 can be a subset of the input parameters of Model 0 (as illustrated). For example, input parameters for the model may include historical measurements, motion detection from a gyroscope, output / trigger from a previous model, previous measurements, other cell / beam measurements, measurements from the RAT, RRC status, atmospheric pressure, direction and distance of travel, orientation, and / or GNSS information. As shown, Model 1 104 can have fewer inputs than Model 0 106. In some implementations, Model 1 104 and Model 0 106 may use different sets of inputs.

[0022] Model 0 106 can be activated based on Model 1 predictions that events of interest might be predicted by Model 0 106. Model 1 104 typically uses less memory and less computational power than Model 0 106. For example, Model 1 104 may have fewer inputs, fewer hidden layers, and simpler outputs than Model 0 106, resulting in less computational requirements and less power consumption.

[0023] In some implementations, more than two stages may be used. For example, three or more models may exist that the UE runs. In some implementations, models may trigger each other serially. In some implementations, more complex models may be operated using triggers from more than one model. In some implementations, if actual measurements are available, the UE is expected to use actual measurements instead of predicted measurements. This may apply to both the final (model 0 106) output and intermediate (e.g., model 1 104) outputs.

[0024] In some implementations, model 1 104 runs continuously. For example, model 1 104 may be triggered by time or system frames, or every n system frames. In some implementations, a threshold defined for the input of model 1 104 may exist, which triggers inference from model 1 104. For example, model 1 104 may not run until the input 'a' reaches the threshold, rather than the UE running model 1 104 continuously.

[0025] In some implementations, Model 0 106 may use synchronization signal blocks (SSBs) and / or channel state information reference signals (CSI-RS) to predict the reference signal received power (RSRP) and reference signal received quality (RSRQ) for each cell and / or each beam. Alternatively or otherwise, Model 0 106 may predict mobility events.

[0026] For the "next model" in the chain (e.g., model 0 106), there may be defined trigger thresholds to determine when the model runs. For example, the output of model 1 104 could be the percentage of the probability that model 0 106 will predict a mobility event or measurement. Model 0 106 can run when its output equals or exceeds the threshold. In some implementations, the model may be trained based on actual measurements and results.

[0027] For example, the UE can collect measurements and process them normally to determine when a mobility event occurs. When a mobility event occurs, the UE can transmit a measurement report to the network. The network can determine whether a handover is triggered. This event and measurement chain can be used to train one or more models in a model. Once the model is trained to a sufficient level of accuracy, it can be used to predict the likelihood of the current situation triggering a mobility event. In some implementations, if a mobility event is likely, model 0106 can predict the measurements to be transmitted to the network to determine whether a handover is triggered.

[0028] Figure 2 Example block diagram 202 illustrates three chained AI / ML models according to some implementation schemes. In the illustrated implementation scheme, model 0 204 can predict actual measurements. For example, model 0 204 can use SSB and / or CSI-RS to predict RSRQ for each cell and / or each beam. Alternatively or in addition, model 0 204 can predict mobility events (which can be a refinement of events predicted by model 1 206, as described below).

[0029] Model 1 206 predicts mobility events. For example, Model 1 206 can predict event A1 (service becomes better than a threshold) or event A2 (service becomes worse than a threshold) or any other event from at least the Ax and Bx series. The output of Model 1 206 can trigger Model 0, and the trigger indicates the predicted event.

[0030] Model 2 208 predicts whether a mobility event is likely to occur. The output of Model 2 208 can trigger Model 1 206. This output trigger can be a binary indication that a mobility event is likely to occur. For example, if Model 2 208 predicts that the probability of a mobility event is greater than a threshold, it can send this output trigger to Model 1 206.

[0031] In some implementations, the model can be trained based on actual measurements and results. For example, the UE can collect measurements and process them normally to determine when a mobility event occurs. When a mobility event occurs, the UE can transmit a measurement report to the network. The network can determine whether a handover is triggered. This event and measurement chain can be used to train one or more models in the model. Once the model is trained to a sufficient level of accuracy, it can be used to predict the probability of the current situation triggering a mobility event and predict measurements. The predicted measurements can be transmitted by the UE to the network node in a measurement report, and the network node can determine whether a handover is triggered based on this measurement report.

[0032] In some implementations, models may be combined into a single model (e.g., Model 2 208 and Model 1 206, or Model 1 206 and Model 0 204). In some implementations, in some cases (e.g., depending on network configuration or UE-specific implementation selection), the output of Model 1 may be sufficient, while in other cases, Model 0 204 may also be triggered. For example, in some cases, predictions of mobility events from Model 1 206 are used, and in other cases, measurement predictions from Model 0 204 may also be desired.

[0033] In some implementations, the output of model 2208 may be minimally processed through model 1206 (or be transparent to model 1206) to directly trigger model 0204. For example, some inputs (e.g., the output of model 2208) may be used by multiple models in the chain. Each model in the chain may be run by the same UE.

[0034] As the diagram illustrates, as the models become more complex, the number of inputs used for each model in the chain may increase sequentially. For example, model 2 (208) receives two inputs, model 1 (206) receives three inputs, and model 0 (204) receives six inputs. Furthermore, models can use triggers from the outputs of previous models as inputs. As shown, in some inputs, lower-complexity models may use a subset of the inputs used by higher-complexity models. In other implementations, models may each use different inputs.

[0035] In some implementations, two or more models are used (e.g., Figure 1 and Figure 2 The model chain can be used to predict various items of interest. The desired predictions can be pre-configured or set by the network.

[0036] In some implementations, Model 1 predicts beam measurements, and Model 0 predicts cell measurements. In some implementations, Model 1 can be trained to predict the optimal beam or multiple beams.

[0037] In some implementations, Model 1 can predict cell measurements. Additionally, Model 0 can predict beam measurements. For example, Model 0 can be trained to predict the optimal beam or multiple beams.

[0038] In some implementations, Model 1 can predict the optimal beam measurement. Furthermore, Model 0 can predict multiple beam measurements.

[0039] In some implementations, multiple models can be used simultaneously (i.e., in parallel) to trigger the next model in the chain. For example, in Figure 2 In this model, multiple models can replace model 2 208, and the models can run in parallel. In some implementations, when the output of any model in the parallel models is triggered, the next model in the chain (e.g., model 1 206) can run. In some implementations, when the output of each model in the parallel models is triggered, the next model in the chain (e.g., model 1 206) can run.

[0040] Figure 3 Example block diagram 302 illustrates a chained AI / ML model with a shared early-stage model according to some implementations. In some implementations, one or more models in the chained model can be used to trigger multiple prediction models. For example, two models or a model chain can be triggered by an early-stage model. This can take advantage of the fact that in some implementations, other models used in the UE may exist in addition to the L2 mobility AI / ML model. One such example could be an L1 layer model for CSI prediction. Therefore, some implementations may use a shared early-stage model (e.g., a shared model for mobility model chains and CSI prediction models).

[0041] In the illustrated implementation scheme, instead of the dedicated early-stage model (i.e., Figure 1 and Figure 2 In this method, another model (e.g., L1 layer CSI prediction model 304) is trained to also trigger an L2 mobility model (e.g., model 1 306). This approach leverages the fact that some L1 models (e.g., CSI prediction model 304) operate at least under RRC connections with a small temporal resolution and are likely always active. Thus, there is minimal overhead (in terms of computational resources) in running the model that triggers the L2 mobility model (e.g., model 1 306).

[0042] For example, in the illustrated embodiment, a Layer 1 (L1) model (e.g., CSI prediction model 304) can be used for CSI prediction. CSI prediction model 304 can receive inputs (e.g., input x and input y) and output a CSI prediction based on the inputs. While the output of CSI prediction model 304 in the illustrated embodiment is a CSI prediction, in other embodiments, the output can be a trigger for causing another model to predict CSI.

[0043] Additionally, CSI prediction model 304 can use input to determine the likelihood of a mobility event. If CSI prediction model 304 determines that the likelihood of a mobility event exceeds a threshold, it can output a trigger to model 1 306, which may be an L2 mobility model.

[0044] Model 1 306 can run when it receives a trigger from CSI Prediction Model 304. Model 1 306 can receive inputs (e.g., input a, input b, and input c). Model 1 306 can use the inputs to predict various variables. These predictions can be measurements and / or more accurate predictions of mobility events. The inputs to Model 1 306 may be different from those of CSI Prediction Model 304, or may include one or more of the same inputs as CSI Prediction Model 304.

[0045] Model 1 306 may transmit a trigger to run Model 0 308 when it determines that the probability of a mobility event exceeds a threshold. In some implementations, Model 1 306 may be omitted, and a trigger from CSI prediction model 304 may run Model 0 308. Model 0 308 may output a mobility inference decision.

[0046] For some implementation schemes, the model inputs (e.g.) Figure 1 , Figure 2 and Figure 3 The models in the model chain may include one or more of the following: Model inputs may include the outputs / triggers of previous models (in the multi-stage model chain). The outputs / triggers of previous models may be the confidence levels of previous models. Model inputs may include previous measurements (in the case of time-based prediction). Previous measurements may be RSRP or RSRQ based on SSB or CSI-RS during the target time window. Model inputs may include other cell / beam measurements (in the case of spatial prediction). In some embodiments, inputs may include historical measurements used for other cell / beam measurements. In some embodiments, inputs may include measurements from other RATs, including 3GPP (e.g., LTE) and / or non-3GPP (e.g., Wi-Fi, Bluetooth, and Ultra Wideband (UWB)). In some embodiments, inputs may include RRC states (e.g., RRC idle, RRC connected, and RRC inactive).

[0047] In some implementations, inputs may include sensor information. Sensor information may include atmospheric pressure, detected movement (e.g., movement detected by a gyroscope), direction and distance of travel (as measured by sensors), orientation / azimuth, GNSS information (position, velocity, and time), and / or proximity to obstacles (as measured by proximity sensors). The model may use one or more of this type of sensor information.

[0048] In some implementations, the early-stage models in the chain may use different inputs than the final-stage models. In some implementations, the early-stage models in the chain may use a subset of the inputs used by the final-stage models. In some implementations, the early-stage models in the chain may use parameters obtained with higher power efficiency. For example, the early-stage models may use sensor information as input, and the later or final-stage models may use actual measurements, outputs from the early-stage models, and sensor information.

[0049] For some implementation schemes, the model output (e.g.) Figure 1 , Figure 2 and Figure 3 The models in the model chain may include one or more of the following. In some embodiments, the output may include whether a mobility-related event may occur. In some embodiments, the output may include which mobility events may occur (e.g., A1 to A6, B1 to B2, etc.). In some embodiments, the output may include measurement predictions. The predicted measurements may be cell-level and / or beam-level. The predicted measurements may be SSB-based and CSI-RS-based, and may be RSRP, RSRQ, and / or signal-to-interference-plus-noise ratio (SINR). The output may also include triggering for the next model in the chain. In some embodiments, the trigger may be associated with a confidence level above a threshold. The output may also include a confidence level.

[0050] In some implementations, the multi-stage model described herein may have an impact on the standard. For example, there may be an impact on the stage 2 description of the method in TS38.300. In some implementations, changes may be made to the UE procedure description in TS 38.331. For example, changes may be included regarding the measurement procedure in section 5.5 to describe the chained model usage method. Additionally, the chained model method may be included in the UE capabilities in TS 38.306. In some implementations, the specification may include the network control description in TS38.331 regarding whether the use of such a model is permitted (e.g., via dedicated RRC signaling at UE resolution, or via, for example, at the cell level via SIB).

[0051] For example, network nodes can be configured with model chains. This can be an enhancement to AI / ML lifecycle management (LCM) model selection. Various implementations are possible, which may include different degrees of network control over how model chains are configured. For example, in some implementations without model linking, the network will select one model for AI / ML mobility. With linking, the network can select two (or more) models and can also signal how models are linked (e.g., “Model 0 ID”, “Model 1 ID”, “Model 2 ID” instead of just “Model ID”). In some implementations, model chains may correspond to model chain IDs.

[0052] In some implementations, the network can signal a confidence level threshold to trigger the next model in the chain. In some implementations, the network can also perform a finer level of control to signal each model not only the confidence level threshold but also the configuration of the input parameters, some of which may be the outputs of previous models in the chain and some new parameters.

[0053] In some implementations, enhancements may be provided to measurement reporting from the UE to the network node. In some implementations, the UE may indicate whether the reported value is an actual measurement or a predicted value. In the case of a predicted value triggering the report, the UE may report information about the model chain (and its triggering) that led to the report.

[0054] In some implementations, the self-optimizing network (SON) can be enhanced. For example, a new report can be defined that reports situations where an early-stage model output triggers a measurement (the output of the final-stage model, e.g., (Model 0)) that exceeds a threshold requiring a measurement report. For example, the UE can transmit a report to a network node indicating a near-failure event predicted by the early-stage model but not confirmed by the later-stage model. In some implementations, the UE can report to a network node when the model does not predict an event, but the actual measurement results in the event.

[0055] Figure 4 An example method 400 performed by a UE according to some implementation schemes is illustrated. As shown, method 400 includes receiving 402 configuration for an AI / ML model chain from a network node, wherein the AI / ML model chain includes multiple models used by the UE, wherein an early-stage model triggers the execution of a final-stage model. Method 400 also includes configuring 404 the early-stage model and the final-stage model according to the configuration, wherein the output from the early-stage model triggers the operation of the final-stage model. Method 400 also includes using the AI / ML model chain to predict 406 measurement or mobility events. Method 400 also includes reporting 408 the prediction to the network node.

[0056] In some implementations, predicting a measurement or mobility event includes: using an early-stage model to predict the confidence level of the occurrence of the mobility event; and when the confidence level exceeds a threshold, transmitting the output from the early-stage model to the final-stage model to trigger the operation of the final-stage model.

[0057] In some implementations, the configuration for AI / ML model chains includes identifiers for two or more models and indicates how the two or more models are linked.

[0058] In some implementations, forecasting includes an indication of whether a mobility-related event is likely to occur, an indication of what type of mobility event is likely to occur, a measurement of the forecast, or a combination thereof.

[0059] In some implementations, the input to the model in the early stages is less than the input to the model in the final stages.

[0060] In some implementations, method 400 also includes configuring the third-stage model to trigger the operation of the earlier-stage model.

[0061] In some implementations, the early-stage model runs continuously, and the final-stage model runs when triggered by the early-stage model.

[0062] In some implementations, the early-stage model is used for Layer 1 prediction and to trigger Layer 2 mobility prediction.

[0063] The embodiments contemplated herein include an apparatus comprising components for performing one or more elements of method 400. This apparatus may be, for example, a UE (such as wireless device 702 as a UE, as described herein).

[0064] The embodiments contemplated herein include one or more non-transitory computer-readable media comprising instructions that, when executed by one or more processors of an electronic device, cause the electronic device to perform one or more elements of method 400. The non-transitory computer-readable medium may be, for example, the memory of a UE (such as memory 706 of a wireless device 702 as a UE, as described herein).

[0065] The embodiments contemplated herein include an apparatus comprising logic components, modules, or circuitry for performing one or more elements of method 400. This apparatus may be, for example, a UE (such as wireless device 702 as a UE, as described herein).

[0066] The embodiments contemplated herein include an apparatus comprising: one or more processors and one or more computer-readable media including instructions that, when executed by the one or more processors, cause the one or more processors to perform one or more elements of method 400. The apparatus may be, for example, a UE (such as wireless device 702 as a UE, as described herein).

[0067] The implementation scheme envisioned herein includes a signal as described in or related to one or more elements of method 400.

[0068] The embodiments contemplated herein include a computer program or computer program product comprising instructions, wherein execution by a processor causes the processor to perform one or more elements of method 400. The processor may be a processor of the UE (such as processor 704 as a wireless device 702 of the UE, as described herein). These instructions may, for example, reside in the processor and / or in the memory of the UE (such as memory 706 as a wireless device 702 of the UE, as described herein).

[0069] Figure 5 An example method 500, performed by a network node according to some implementation schemes, is illustrated. Method 500 includes generating 502 a configuration for an AI / ML model chain, wherein the AI / ML model chain includes multiple models used by the UE, wherein early-stage models trigger the execution of final-stage models. Method 500 also includes transmitting 504 the configuration for the AI / ML model chain to the user equipment. Method 500 also includes receiving 506 predictions about measurement or mobility events from the UE based on the AI / ML model chain.

[0070] In some implementations, the AI / ML model chain uses an early-stage model to predict the confidence level of a mobility event; and when the confidence level exceeds a threshold, the early-stage model delivery triggers the final-stage model to run.

[0071] In some implementations, the configuration for AI / ML model chains includes identifiers for two or more models and instructions on how those two or more models are linked.

[0072] In some implementations, forecasting includes an indication of whether a mobility-related event is likely to occur, an indication of what type of mobility event is likely to occur, a measurement of the forecast, or a combination thereof.

[0073] In some implementations, the input to the model in the early stages is less than the input to the model in the final stages.

[0074] In some implementations, the configuration for the AI / ML model chain also includes a third-stage model for triggering operations of the earlier-stage models.

[0075] In some implementations, the early-stage model runs continuously, and the final-stage model runs when triggered by the early-stage model.

[0076] The embodiments contemplated herein include an apparatus comprising components for performing one or more elements of method 500. This apparatus may be, for example, a base station (such as network device 718 as a base station, as described herein).

[0077] The embodiments contemplated herein include one or more non-transitory computer-readable media comprising instructions that, when executed by one or more processors of an electronic device, cause the electronic device to perform one or more elements of method 500. The non-transitory computer-readable medium may be, for example, the memory of a base station (such as memory 722 of a network device 718 serving as a base station, as described herein).

[0078] The embodiments contemplated herein include an apparatus comprising logic components, modules, or circuitry for performing one or more elements of method 500. This apparatus may be, for example, an apparatus for a base station (such as network device 718 as a base station, as described herein).

[0079] The embodiments contemplated herein include an apparatus comprising: one or more processors and one or more computer-readable media including instructions that, when executed by the one or more processors, cause the one or more processors to perform one or more elements of method 500. The apparatus may be, for example, an apparatus for a base station (such as network device 718 as a base station, as described herein).

[0080] The implementation scheme envisioned herein includes a signal as described in or related to one or more elements of method 500.

[0081] The embodiments contemplated herein include a computer program or computer program product comprising instructions, wherein execution by a processing element causes the processing element to perform one or more elements of method 500. The processor may be a processor of a base station (such as processor 720 of network device 718 as a base station, as described herein). These instructions may, for example, reside in the processor and / or in the memory of the base station (such as memory 722 of network device 718 as a base station, as described herein).

[0082] Figure 6An example architecture of a wireless communication system 600 according to an embodiment disclosed herein is illustrated. The following description is provided for an example wireless communication system 600 operating in conjunction with LTE system standards and / or 5G or NR system standards provided by 3GPP technical specifications.

[0083] like Figure 6 As shown, the wireless communication system 600 includes UE 602 and UE 604 (but any number of UEs may be used). In this example, UE 602 and UE 604 are exemplified as smartphones (e.g., handheld touchscreen mobile computing devices capable of connecting to one or more cellular networks), but may also include any mobile or non-mobile computing device configured for wireless communication.

[0084] UE 602 and UE 604 can be configured to communicatively couple with RAN 606. In an embodiment, RAN 606 can be NG-RAN, E-UTRAN, etc. UE 602 and UE 604 utilize connections (or channels) with RAN 606 (shown as connection 608 and connection 610, respectively), each of these connections including a physical communication interface. RAN 606 may include one or more base stations (such as base station 612 and base station 614) implementing connection 608 and connection 610.

[0085] In this example, Connection 608 and Connection 610 are the air interfaces that enable this communication coupling and are compliant with the RAT used by RAN 606, such as LTE and / or NR.

[0086] In some implementations, UE 602 and UE 604 may also exchange communication data directly via sidelink interface 616. UE 604 is shown configured to access an access point (shown as AP 618) via connection 620. For example, connection 620 may include a local wireless connection, such as a connection conforming to any IEEE 802.11 protocol, while AP 618 may include Wi-Fi. ® Router. In this example, AP 618 may connect to another network (e.g., the Internet) without using CN 624.

[0087] In the implementation, UE 602 and UE 604 may be configured to communicate with each other or with base station 612 and / or base station 614 on a multi-carrier communication channel using orthogonal frequency division multiplexing (OFDM) communication signals according to various communication technologies, such as, but not limited to, orthogonal frequency division multiple access (OFDMA) communication technology (e.g., for downlink communication) or single-carrier frequency division multiple access (SC-FDMA) communication technology (e.g., for uplink and ProSe or sidelink communication) , but the scope of the implementation is not limited in this respect. The OFDM signal may include multiple orthogonal subcarriers.

[0088] In some implementations, all or some of the base stations in base station 612 or base station 614 may be implemented as one or more software entities running on a server computer as part of a virtual network. Furthermore, or in other implementations, base station 612 or base station 614 may be configured to communicate with each other via interface 622. In implementations where the wireless communication system 600 is an LTE system (e.g., when CN 624 is an EPC), interface 622 may be an X2 interface. This X2 interface may be defined between two or more base stations (e.g., two or more eNBs, etc.) connected to the EPC and / or between two eNBs connected to the EPC. In implementations where the wireless communication system 600 is an NR system (e.g., when CN 624 is a 5GC), interface 622 may be an Xn interface. The Xn interface is defined between two or more base stations (e.g., two or more gNBs, etc.) connected to the 5GC, between a base station 612 (e.g., a gNB) connected to the 5GC and an eNB, and / or between two eNBs connected to the 5GC (e.g., CN 624).

[0089] RAN 606 is shown communicatively coupled to CN 624. CN 624 may include one or more network elements 626 configured to provide various data and telecommunications services to customers / subscribers (e.g., users of UE 602 and UE 604) connected to CN 624 via RAN 606. Components of CN 624 may be implemented in a single physical device or a separate physical device including components for reading and executing instructions from machine-readable or computer-readable media (e.g., non-transitory machine-readable storage media).

[0090] In the implementation scheme, CN 624 may be an EPC, and RAN 606 may be connected to CN 624 via S1 interface 628. In the implementation scheme, S1 interface 628 may be divided into two parts: an S1 user plane (S1-U) interface, which carries service data between base station 612 or base station 614 and the service gateway (S-GW); and an S1-MME interface, which is the signaling interface between base station 612 or base station 614 and the mobility management entity (MME).

[0091] In the implementation scheme, CN 624 may be a 5GC, and RAN 606 may be connected to CN 624 via NG interface 628. In the implementation scheme, NG interface 628 may be divided into two parts: an NG user plane (NG-U) interface, which carries service data between base station 612 or base station 614 and user plane function (UPF); and an S1 control plane (NG-C) interface, which is the signaling interface between base station 612 or base station 614 and access and mobility management function (AMF).

[0092] Generally, application server 630 can be a component that provides Internet Protocol (IP) carried resources (e.g., packet-switched data services) for use with CN 624. Application server 630 can also be configured to support one or more communication services (e.g., VoIP sessions, group communication sessions, etc.) for UE 602 and UE 604 via CN 624. Application server 630 can communicate with CN 624 via IP communication interface 632.

[0093] Figure 7 A system 700 for performing signaling transfer 734 between a wireless device 702 and a network device 718 according to an embodiment disclosed herein is illustrated. System 700 may be part of a wireless communication system as described herein. Wireless device 702 may be, for example, a UE (User Equipment) in a wireless communication system. Network device 718 may be, for example, a base station (e.g., an eNB or gNB) in a wireless communication system.

[0094] Wireless device 702 may include one or more processors 704. Processor 704 is executable instructions that cause various operations of wireless device 702 to be performed as described herein. Processor 704 may include one or more baseband processors, which are implemented using, for example, a central processing unit (CPU), digital signal processor (DSP), application-specific integrated circuit (ASIC), controller, field-programmable gate array (FPGA) device, another hardware device, firmware device, or any combination thereof configured to perform the operations described herein.

[0095] Wireless device 702 may include memory 706. Memory 706 may be a non-transitory computer-readable storage medium that stores instructions 708, which may include instructions executable, for example, by processor 704. Instructions 708 may also be referred to as program code or a computer program. Memory 706 may also store data used by processor 704 and results calculated by the processor.

[0096] Wireless device 702 may include one or more transceivers 710, which may include radio frequency (RF) transmitter circuitry and / or receiver circuitry that use antenna 712 of wireless device 702 to facilitate signaling to and / or from wireless device 702 and other devices (e.g., network device 718) according to a corresponding RAT (e.g., signaling transmission 734).

[0097] Wireless device 702 may include one or more antennas 712 (e.g., one, two, four or more). In embodiments with multiple antennas 712, wireless device 702 can fully utilize the spatial diversity of such multiple antennas 712 to transmit and / or receive multiple different data streams on the same time-frequency resources. This behavior may be referred to as, for example, multiple-input multiple-output (MIMO) behavior (referring to multiple antennas used at each of the transmitting and receiving devices to implement this aspect). MIMO transmission by wireless device 702 can be implemented according to pre-decoding (or digital beamforming) applied to wireless device 702, which multiplexes the data streams among antennas 712 based on known or assumed channel characteristics, such that each data stream is received with appropriate signal strength relative to the others at a desired location in the spatial domain (e.g., the location of the receiver associated with that data stream). Some implementations may use a single-user MIMO (SU-MIMO) approach (where all data streams are directed to a single receiver) and / or a multi-user MIMO (MU-MIMO) approach (where individual data streams may be directed to individual (different) receivers at different locations in the airspace).

[0098] In some implementations with multiple antennas, the wireless device 702 can implement analog beamforming technology, whereby the phase of the signal transmitted by the antenna 712 is relatively adjusted so that the (joint) transmission of the antenna 712 can be directed (this is sometimes referred to as beam control).

[0099] Wireless device 702 may include one or more interfaces 714. Interfaces 714 can be used to provide input to or output to wireless device 702. For example, wireless device 702 as a UE may include interfaces 714, such as a microphone, speaker, touchscreen, buttons, etc., to allow input and / or output from a user of the UE to the UE. Other interfaces of such UEs may consist of transmitters, receivers, and other circuitry that allow the UE to communicate with other devices (e.g., in addition to the transceiver 710 / antenna 712 already described), and may be based on known protocols (e.g., Wi-Fi). ® and Bluetooth ® (etc.) to perform the operation.

[0100] Wireless device 702 may include model chain module 716. Model chain module 716 may be implemented via hardware, software, or a combination thereof. For example, model chain module 716 may be implemented as a processor, circuitry, and / or instructions 708 stored in memory 706 and executed by processor 704. In some examples, model chain module 716 may be integrated within processor 704 and / or transceiver 710. For example, model chain module 716 may be implemented by a combination of software components and hardware components (e.g., logic gates and circuitry) within processor 704 or transceiver 710 (e.g., executed by a DSP or general-purpose processor).

[0101] The model chain module 716 can be used in various aspects of this disclosure, for example, Figures 1 to 6 All aspects. The model chain module 716 is configured to configure the model chain and use the model chain to perform predictions.

[0102] Network device 718 may include one or more processors 720. Processor 720 is executable instructions to perform various operations of network device 718 as described herein. Processor 720 may include one or more baseband processors, which are implemented using, for example, a CPU, DSP, ASIC, controller, FPGA device, another hardware device, firmware device, or any combination thereof configured to perform the operations described herein.

[0103] Network device 718 may include memory 722. Memory 722 may be a non-transitory computer-readable storage medium that stores instructions 724, which may include, for example, instructions executed by processor 720. Instructions 724 may also be referred to as program code or a computer program. Memory 722 may also store data used by processor 720 and results calculated by the processor.

[0104] Network device 718 may include one or more transceivers 726, which may include RF transmitter circuitry and / or receiver circuitry that use antenna 728 of network device 718 to facilitate signaling transmission to and / or from network device 718 and other devices (e.g., signaling transmission 734) in accordance with corresponding RAT.

[0105] Network device 718 may include one or more antennas 728 (e.g., one, two, four or more). In embodiments having multiple antennas 728, network device 718 may perform MIMO, digital beamforming, analog beamforming, beam control, etc., as described.

[0106] Network device 718 may include one or more interfaces 730. Interfaces 730 may be used to provide input to or output to network device 718. For example, network device 718 as a base station may include interfaces 730 consisting of transmitters, receivers, and other circuitry (e.g., in addition to the transceiver 726 / antenna 728 described), which enable the base station to communicate with other equipment in the core network and / or enable the base station to communicate with external networks, computers, databases, etc., for the purpose of performing operations, management, and maintenance of the base station or other equipment operatively connected to the base station.

[0107] Network device 718 may include model chain configuration module 732. Model chain configuration module 732 may be implemented via hardware, software, or a combination thereof. For example, model chain configuration module 732 may be implemented as a processor, circuitry, and / or instructions 724 stored in memory 722 and executed by processor 720. In some examples, model chain configuration module 732 may be integrated within processor 720 and / or transceiver 726. For example, model chain configuration module 732 may be implemented via a combination of software components (e.g., executed by a DSP or general-purpose processor) and hardware components (e.g., logic gates and circuitry) within processor 720 or transceiver 726.

[0108] The model chain configuration module 732 can be used in various aspects of this disclosure, for example, Figures 1 to 6 All aspects. The model chain configuration module 732 is configured to generate the configuration for the model chain to be transmitted to the wireless device 702.

[0109] For one or more embodiments, at least one of the components illustrated in one or more of the foregoing figures may be configured to perform one or more operations, techniques, processes, and / or methods as described herein. For example, a baseband processor as described herein in conjunction with one or more of the foregoing figures may be configured to operate according to one or more of the examples illustrated herein. Similarly, circuitry associated with a UE, base station, network element, etc., as described above in conjunction with one or more of the foregoing figures may be configured to operate according to one or more of the examples illustrated herein.

[0110] Unless otherwise explicitly stated, any of the embodiments described above may be combined with any other embodiment (or combination of embodiments). The foregoing description of one or more specific embodiments provides illustrative and descriptive information, but is not intended to be exhaustive or to limit the scope of the embodiments to the precise form disclosed. In light of the teachings above, modifications and variations are possible, or modifications and variations may be derived from practice with various embodiments.

[0111] Implementations and specific embodiments of the systems and methods described herein may include various operations embodied in machine-executable instructions to be executed by a computer system. The computer system may include one or more general-purpose or special-purpose computers (or other electronic devices). The computer system may include hardware components, including specific logical parts for performing the operations; or may include a combination of hardware, software, and / or firmware.

[0112] It should be recognized that the systems described herein include descriptions of specific implementations. These implementations may be combined into a single system, partially integrated into other systems, divided into multiple systems, or otherwise partitioned or combined. Furthermore, it is conceivable to use parameters, attributes, aspects, etc., of one implementation in one implementation. For clarity, these parameters, attributes, aspects, etc., are described only in one or more implementations, and it should be recognized that, unless expressly stated herein, these parameters, attributes, aspects, etc., may be combined with or substituted for parameters, attributes, aspects, etc., of another implementation.

[0113] As is widely recognized, the use of personally identifiable information should comply with privacy policies and practices that are generally accepted to meet or exceed industry or governmental requirements for protecting user privacy. Specifically, personally identifiable information data should be managed and processed to minimize the risk of unintentional or unauthorized access or use, and the nature of authorized use should be clearly explained to users.

[0114] Although the foregoing has been described in considerable detail for clarity, it will be apparent that certain changes and modifications can be made without departing from the principles of the invention. It should be noted that there are many alternative ways to implement both the processes and apparatus described herein. Therefore, embodiments of the invention should be considered illustrative rather than restrictive, and this specification is not limited to the details given herein, but can be modified within the scope and equivalents of the appended claims.

Claims

1. A method performed by a user equipment (UE), the method comprising: Receive configuration from network nodes for an artificial intelligence (AI) / machine learning (ML) model chain, wherein the AI / ML model chain includes multiple models used by the UE, wherein early-stage models trigger the execution of final-stage models; The early-stage model and the final-stage model are configured according to the configuration, wherein the output from the early-stage model triggers the operation of the final-stage model; Use the AI / ML model chain to predict measurement or mobility events; as well as Report the prediction to the network nodes.

2. The method of claim 1, wherein predicting the measurement or the mobility event comprises: The early-stage model is used to predict the confidence level of the occurrence of the mobility event; as well as When the confidence level exceeds a threshold, the output from the earlier stage model is transmitted to the final stage model to trigger the operation of the final stage model.

3. The method of claim 1, wherein the configuration for the AI / ML model chain includes identifiers for two or more models and indicates how the two or more models are linked.

4. The method of claim 1, wherein the prediction includes an indication of whether a mobility-related event may occur, an indication of what type of mobility event may occur, a measurement of the prediction, or a combination thereof.

5. The method of claim 1, wherein the input of the early stage model is less than the input of the final stage model.

6. The method of claim 1, further comprising configuring the third-stage model to trigger the operation of the earlier-stage model.

7. The method of claim 1, wherein the early stage model runs continuously, and the final stage model runs when triggered by the early stage model.

8. The method of claim 1, wherein the early stage model is used for layer 1 prediction and to trigger layer 2 mobility prediction.

9. A method executed by a network node, the method comprising: Generate a configuration for an artificial intelligence (AI) / machine learning (ML) model chain, wherein the AI / ML model chain includes multiple models used by the user experience (UE), wherein an early-stage model triggers the execution of a final-stage model; Transmit the configuration for the AI / ML model chain to the user equipment; and Based on the AI / ML model chain, predictions about measurement or mobility events are received from the UE.

10. The method of claim 9, wherein the AI / ML model chain: The early-stage model is used to predict the confidence level of the mobility event; and When the confidence level exceeds the threshold, the early stage model delivery is triggered to run the final stage model.

11. The method of claim 9, wherein the configuration for the AI / ML model chain includes identifiers for two or more models and indications of how the two or more models are linked.

12. The method of claim 9, wherein the prediction includes an indication of whether a mobility-related event may occur, an indication of what type of mobility event may occur, a measurement of the prediction, or a combination thereof.

13. The method of claim 9, wherein the input to the early stage model is less than the input to the final stage model.

14. The method of claim 9, wherein the configuration for the AI / ML model chain further includes a third-stage model for triggering operations of the early-stage model.

15. The method of claim 9, wherein the early stage model operates continuously, and the final stage model operates when triggered by the early stage model.

16. The method of claim 9, wherein the early stage model is used for layer 1 prediction and to trigger layer 2 mobility prediction.

17. A baseband processor for user equipment (UE), the baseband processor being configured to cause the UE to: Receive configuration from network nodes for an artificial intelligence (AI) / machine learning (ML) model chain, wherein the AI / ML model chain includes multiple models used by the UE, wherein early-stage models trigger the execution of final-stage models; The early-stage model and the final-stage model are configured according to the configuration, wherein the output from the early-stage model triggers the operation of the final-stage model; Use the AI / ML model chain to predict measurement or mobility events; and Report the prediction to the network nodes.

18. The baseband processor of claim 17, wherein predicting the measurement or the mobility event comprises: The early-stage AI / ML model is used to predict the confidence level of the occurrence of the mobility event; as well as When the confidence level exceeds a threshold, the output from the earlier stage model is transmitted to the final stage model to trigger the operation of the final stage model.

19. The baseband processor of claim 17, wherein the configuration for the AI / ML model chain includes identifiers for two or more models and indicates how the two or more models are linked.

20. The baseband processor of claim 17, wherein the prediction includes an indication of whether a mobility-related event may occur, an indication of what type of mobility event may occur, a measurement of the prediction, or a combination thereof.

21. The baseband processor of claim 17, wherein the input to the early stage model is less than the input to the final stage model.

22. The baseband processor of claim 17, wherein the baseband processor is further configured to cause the UE to configure the third-stage model to trigger the operation of the earlier-stage model.

23. The baseband processor of claim 17, wherein the early stage model operates continuously, and the final stage model operates when triggered by the early stage model.

24. The baseband processor of claim 17, wherein the early stage model is used for layer 1 prediction and to trigger layer 2 mobility prediction.

25. An apparatus comprising components for performing the method according to any one of claims 1 to 16.

26. A computer-readable medium comprising instructions that, when executed by one or more processors of an electronic device, cause the electronic device to perform the method according to any one of claims 1 to 16.

27. An apparatus comprising a logic component, module, or circuit for performing the method according to any one of claims 1 to 16.

Images