Method and System for UE-Assisted Data Collection for Mobility Prediction

Abstract

Systems and methods for user equipment (UE)-assisted data collection for mobility prediction are disclosed herein. In one embodiment, a method for UE-assisted data collection for mobility prediction performed by a UE includes: receiving, from a network node of a cellular communication system, a UE trajectory prediction model for predicting a UE trajectory. The method further includes: executing the UE trajectory prediction model to generate a predicted trajectory of the UE; comparing the actual UE trajectory with the predicted UE trajectory; and sending, to the network node of the cellular communication system, a result of the comparison between the actual UE trajectory and the predicted UE trajectory. In this way, the mobility prediction performance is improved.

Description

[0001] Related Applications

[0002] This application claims the benefit of U.S. Provisional Patent Application Ser. No. 62 / 880,464, filed Jul. 30, 2019, the entire disclosure of which is hereby incorporated by reference herein. Field of the Disclosure

[0003] The present disclosure relates to mobility prediction in cellular communication systems. Background Art

[0004] The 3rd Generation Partnership Project (3GPP) Network Data Analytics Function (NWDAF) hosts several services around network data analytics. For example, a Service Function (SF) can subscribe to the NWDAF service for providing network slice congestion events. 3GPP Technical Specification (TS) 23.288 contains a list of the currently available services in NWDAF. The specification of NWDAF is being developed in 3GPP, and it is expected that multiple services will be added. Some of these services may be standardized, and some may remain proprietary.

[0005] Ericsson Research has developed an Artificial Intelligence (AI) model in NWDAF that can predict the next Radio Base Station (RBS) for a mobile user device or User Equipment (UE). This model can be used, for example, to optimize paging or to optimize the selection of User Plane Function (UPF) instances to be used. See, for example, International Patent Application PCT / SE2019 / 050235, filed Mar. 15, 2019.

[0006] UE trajectory prediction models require data for training and retraining. Most of this data comes from the Access Management Function (AMF) in the 5th Generation (5G) Core (5GC) network or from the Mobility Management Entity (MME) in the 4th Generation (4G) network, and either the Access Management Function (AMF) or the Mobility Management Entity (MME) has obtained its information from the UE and the Radio Access Network (RAN). This operation is shown in Figure 1 in.

[0007] Figure 1 shows the operation of a traditional NWDAF trajectory prediction service. As shown, an Operation, Administration, and Maintenance (OAM) node subscribes to the NWDAF's trajectory prediction service (step 100) and provides Key Performance Indicators (KPIs) as part of the subscription request. The NWDAF then subscribes to events from the AMF (step 102). The NWDAF then sends a response to the subscription request to the OAM (step 104). In Figure 1In the example shown, the SF also sends a request to subscribe to the NWDAF's trajectory prediction service (step 106) and receives a response to the request (step 108). This interaction is optional and is intended to show that multiple entities can subscribe to the NWDAF's trajectory prediction service.

[0008] When the UE and other mobile devices change their locations, this triggers some communication with a New Radio (NR) base station (gNB) (step 110). Some of these communications trigger the detection of events (step 112) and the generation of event notification messages for the AMF (step 114). The notification of these events is then sent from the AMF to the NWDAF (step 116), where each event contains information about the UE's RBS transition. Each event includes at least a timestamp, a unique UE identifier, and an identifier of the new RBS. Even though additional information (such as the previous RBS and the time elapsed at the previous RBS) can also be deduced by the NWDAF from previously received events, the event may include such information.

[0009] The NWDAF runs a service for UE trajectory prediction. This service uses the events received from the AMF to perform trajectory prediction (step 118). In Figure 1 the example shown, two SFs have subscribed to the NWDAF trajectory prediction service to receive notifications about the service status periodically or to receive the actual trajectory prediction when the service is ready (steps 120, 122).

[0010] Each prediction about which RBS the UE will go to next will have a calculated accuracy. By comparing the prediction (e.g., the predicted trajectory) with the information in the transition events from the AMF (e.g., the actual trajectory), the prediction accuracy can be calculated or measured. A typical KPI is the percentage of correctly predicted transitions. This KPI will change over time. When the model is still untrained, the performance will be low. But for various reasons, the KPI of a trained model can also change over time. For example, when too little retraining data is provided. In our use case, too little data is provided when too many UEs generate too few events over too long a period.

[0011] When the performance of the prediction service is higher than a given KPI (or given KPIs), the present disclosure defines the prediction service as "ready". The minimum KPI is specified by the SF requesting the prediction service. In Figure 1 the example shown, the OAM specifies the KPI it needs in the subscription message (step 100), and the SF specifies the KPI it needs in its subscription message (step 106). As in Figure 1As can be seen, KPIs can be provided by the OAM, by the subscribed users of the service, or by both. In the latter case, the OAM can provide the minimum KPIs. When the service becomes "ready" or "not ready", the subscriber is notified (steps 120 - 122). If the subscriber receives a notification that the service has become ready (meaning that the predicted service performance meets the minimum requirements of the subscriber), the notification can also include a trajectory prediction.

[0012] Regarding the tracking area and the idle mode, when the UE is not receiving or sending any data, the network commands the UE to enter the idle mode. This is a sleep mode to save battery life. In the idle mode, the signaling between the UE and the network is kept to a minimum. However, for example, when there is incoming downlink data towards the UE, the network still needs to be able to reach the UE. This is the case when the so-called paging process arrives. The AMF commands the RBS that last saw the UE to broadcast the message "UE, are you there?". If the UE does not reply, the AMF commands the neighboring RBSs to broadcast the message. If no reply is still received, additional neighbors are queried until the UE replies.

[0013] Without any additional measures, the above paging process is inefficient. In the worst case, all RBSs in the network send broadcast messages to find the UE. To improve efficiency, the concept of a tracking area is used. Each RBS or even each cell of each RBS periodically sends a broadcast message with a tracking area code (TAC). Along with this code, the broadcast also includes the network identity of the operator. The TAC combined with the network identity forms a tracking area identity (TAI). Multiple cells and multiple RBSs can broadcast the same TAI. For each tracking area (TA), there is a unique TAI. What TAI to broadcast is configured once and cannot be changed during operation.

[0014] Each UE is configured with a list of TAIs. The UE receives this list from the AMF during initial registration (initial attachment), and the AMF can change the UE's TAI list over time. When an idle UE moves within its TAI list, it does not need to contact the network. The only exception is that it sends a "I'm still here" message at regular intervals (usually once an hour). But when the UE moves outside its TAI list, it must contact the network to request an updated TAI list. These messages are called "tracking area update" (TAU) messages in 4G or "periodic registration area update" (RAU) messages in 5G. In this document, the terms "tracking area" and "registration area" are used interchangeably. See 3GPP TS 23.501 and 3GPP TS23.401 for the definitions.

[0015] When determining the scale of the network, a balance needs to be found:

[0016] · A large geographical area spanned by the TAI list will result in few TAU messages but will result in many paging messages.

[0017] · A small geographical area spanned by the TAI list will result in few paging messages but will result in many TAU messages.

[0018] There is currently a challenge (or challenges). For example, the AMF can only send mobility events when it has in turn received information from the UE via the RAN. This includes RBS handover, TAU / RAU, or service request (see 3GPP TS 23.502, section 4.2.3). For the UE trajectory prediction model of the NWDAF to perform well, it is important to receive sufficient events regarding RBS transitions. This is contradictory to the current design of 4G / 5G networks where mobility handling signaling is kept to a minimum. In the current design, when the UE is in the idle mode and moves within its tracking area list, there are no RBS handover events. The only events initiated by an idle UE are periodic TAU / RAU, which are completed once per hour in the default configuration. This means that the NWDAF will not receive all RBS transitions of the idle UE. This will in turn lead to poor performance of the NWDAF trajectory model in predicting the next RBS of the UE.

[0019] Therefore, there is a need for systems and methods to address the aforementioned challenges of traditional solutions for UE trajectory prediction. Summary of the Invention

[0020] Systems and methods for UE-assisted data collection for mobility prediction are disclosed herein. In one embodiment, a method for UE-assisted data collection for mobility prediction performed by a UE includes: receiving, from a network node of a cellular communication system, a UE trajectory prediction model for predicting the UE trajectory. The method further includes: executing the UE trajectory prediction model to generate a predicted trajectory of the UE; comparing the actual UE trajectory with the predicted trajectory of the UE; and sending the result of the comparison of the actual UE trajectory and the predicted trajectory of the UE to the network node of the cellular communication system. In this way, the mobility prediction performance is improved.

[0021] In one embodiment, sending the result of the comparison of the actual UE trajectory and the predicted trajectory of the UE includes: sending a transition event or a trajectory event. In another embodiment, sending the result of the comparison of the actual UE trajectory and the predicted trajectory of the UE includes: sending error data describing the difference between the actual UE trajectory and the predicted trajectory of the UE. In one embodiment, the method further includes: sending the result of the comparison of the actual UE trajectory and the predicted trajectory of the UE to another UE.

[0022] In one embodiment, the method further includes: training or retraining a UE trajectory model after comparing an actual UE trajectory with a predicted trajectory of the UE. In one embodiment, sending the result of the comparison between the actual UE trajectory and the predicted trajectory of the UE includes: uploading the trained or retrained UE trajectory prediction model. In one embodiment, the method further includes: sending the trained or retrained UE trajectory prediction model to another UE.

[0023] In one embodiment, receiving a UE trajectory prediction model includes: receiving the UE trajectory prediction model from a Network Data Analytics Function (NWDAF). In another embodiment, receiving a UE trajectory prediction model includes: receiving the UE trajectory prediction model from a Management Data Analytics Function, a Non-Real-Time Intelligent Controller, or a Near-Real-Time Intelligent Controller.

[0024] In one embodiment, sending the result of the comparison includes: sending the result of the comparison to the NWDAF. In another embodiment, sending the result of the comparison includes: sending the result of the comparison to a Management Data Analytics Function, a Non-Real-Time Intelligent Controller, or a Near-Real-Time Intelligent Controller.

[0025] A corresponding embodiment of a UE is also disclosed. In one embodiment, a UE for UE-assisted data collection for mobility prediction is adapted to: receive a UE trajectory prediction model for predicting a UE trajectory from a network node of a cellular communication system. The UE is further adapted to: execute the UE trajectory prediction model to generate a predicted trajectory of the UE; compare the actual UE trajectory with the predicted trajectory of the UE; and send the result of the comparison between the actual UE trajectory and the predicted trajectory of the UE to a network node of the cellular communication system.

[0026] In another embodiment, a UE for UE-assisted data collection for mobility prediction includes: one or more transmitters; one or more receivers; and a processing circuit associated with the one or more transmitters and the one or more receivers. The processing circuit is configured to: cause the UE to receive a UE trajectory prediction model for predicting a UE trajectory from a network node of a cellular communication system. The processing circuit is further configured to: cause the UE to execute the UE trajectory prediction model to generate a predicted trajectory of the UE; compare the actual UE trajectory with the predicted trajectory of the UE; and send the result of the comparison between the actual UE trajectory and the predicted trajectory of the UE to a network node of the cellular communication system.

[0027] Embodiments of a method for UE-assisted data collection for mobility prediction performed by a cellular communication system are also disclosed. In one embodiment, a method for UE-assisted data collection for mobility prediction performed by a cellular communication system includes downloading a UE trajectory prediction model for predicting the trajectory of a target UE to the target UE. The method further includes receiving, from the target UE, information related to training or retraining the UE trajectory prediction model; and using the received information to train or retrain the UE trajectory prediction model.

[0028] In one embodiment, the method further includes updating the UE trajectory prediction model in the target UE. In one embodiment, updating the UE trajectory prediction model in the target UE includes downloading the trained or retrained UE trajectory prediction model to the target UE. In another embodiment, updating the UE trajectory prediction model in the target UE includes downloading updated parameters used by the UE trajectory prediction model in the target UE to the target UE.

[0029] In one embodiment, the method further includes identifying the target UE before downloading the UE trajectory prediction model to the target UE. In one embodiment, identifying the target UE includes receiving a plurality of radio base station (RBS) handover events associated with at least one UE; determining that an RBS handover event of a first UE involves a handover from a first RBS to a second RBS that is not a neighbor of the first RBS; and identifying the first UE as the target for downloading the UE trajectory prediction model.

[0030] In another embodiment, identifying the target UE includes determining that an RBS handover event of each UE in a plurality of UEs involves a handover from a first RBS to a second RBS that is not a neighbor of the first RBS; identifying the plurality of UEs as potential target UEs for downloading respective UE trajectory prediction models; and selecting a subset of the potential target UEs as the target UEs for downloading the respective UE trajectory prediction models, the subset of the potential target UEs including the identified target UE. In one embodiment, selecting a subset of the potential target UEs includes selecting a subset of the potential target UEs based on at least one parameter. In one embodiment, the at least one parameter includes the presence of UEs in an area or region where the performance of the UE trajectory prediction model is below a threshold, UE class or type, UE speed, type of UE movement, UE behavior, radio access technology (RAT) or frequency band supported by the UE, and / or UE battery capacity.

[0031] In one embodiment, at least some of the steps of the method are performed by an NWDAF or an operation, administration, and maintenance (OAM) node.

[0032] In one embodiment, identifying the target UE includes receiving an identification of the target UE from an OAM node.

[0033] Also disclosed is a corresponding embodiment of a system for UE-assisted data collection for mobility prediction. In one embodiment, the system includes: at least one network node for a cellular communication system. The at least one network node is adapted to: download a UE trajectory prediction model for predicting the trajectory of a target UE to the target UE. The at least one network node is further adapted to: receive information related to training or retraining the UE trajectory prediction model from the target UE; and use the received information to train or retrain the UE trajectory prediction model.

[0034] In another embodiment, a method for UE-assisted data collection for mobility prediction performed by a network node includes: receiving a plurality of RBS handover events associated with at least one UE; and identifying a target UE from which to obtain information for training or retraining a corresponding UE trajectory prediction model based on the plurality of RBS handover events.

[0035] In one embodiment, identifying the target UE includes: determining, from among the plurality of RBS handover events, that the RBS handover event of a first UE involves a handover from a first RBS to a second RBS that is not a neighbor of the first RBS; and identifying the first UE as the target UE.

[0036] In one embodiment, identifying the target UE includes: determining, from among the plurality of RBS handover events, that the RBS handover event of each UE among the plurality of UEs involves a handover from a first RBS to a second RBS that is not a neighbor of the first RBS. Identifying the target UE further includes: identifying the plurality of UEs as potential target UEs for downloading a corresponding UE trajectory prediction model; and selecting a subset of the potential target UEs as the target UEs for downloading a corresponding UE trajectory prediction model, the subset of the potential target UEs including the identified target UE. In one embodiment, the subset of the UEs is selected based on one or more parameters. In one embodiment, the one or more parameters include: the presence of UEs in an area or region where the performance of the UE trajectory prediction model is below a threshold, UE class or type, UE speed, type of UE movement, UE behavior, supported RAT or frequency band, and / or UE battery capacity.

[0037] In one embodiment, the network node includes an NWDAF or an OAM node.

[0038] Also disclosed is a corresponding embodiment of a network node for UE-assisted data collection for mobility prediction. In one embodiment, a network node for UE-assisted data collection for mobility prediction is adapted to: receive a plurality of RBS handover events associated with at least one UE; and identify a target UE from which to obtain information for training or retraining a corresponding UE trajectory prediction model based on the plurality of RBS handover events.

[0039] In another embodiment, a network node for UE-assisted data collection for mobility prediction includes processing circuitry configured to cause the network node to: receive a plurality of RBS transition events associated with at least one UE; and identify, based on the plurality of RBS transition events, a target UE from which to obtain information for training or retraining a corresponding UE trajectory prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings, which are incorporated in and constitute a part of this specification, illustrate several aspects of the disclosure and, together with the description, serve to explain the principles of the disclosure.

[0041] Figure 1 Illustrates the operation of a conventional Network Data Analytics Function (NWDAF) User Equipment (UE) trajectory prediction service;

[0042] Figure 2 Illustrates an example of a cellular communication system in which embodiments of the disclosure may be implemented;

[0043] Figure 3 and 4 Illustrates an exemplary embodiment in which the cellular communication system is a Fifth Generation (5G) System (5GS);

[0044] Figure 5A and 5B Illustrates a process for UE-assisted data collection for mobility prediction according to some embodiments of the disclosure;

[0045] Figure 6A and 6B Illustrates a process for UE-assisted data collection for mobility prediction according to some other embodiments of the disclosure;

[0046] Figure 7A and 7B Illustrates a process according to some embodiments of the disclosure by which the system can detect that the performance of the UE trajectory prediction service is not good enough and take measures to increase the number of base station transition events available for training or retraining the UE trajectory prediction model;

[0047] Figure 8 Illustrates a process for identifying a target UE from which to obtain auxiliary information for training or retraining a corresponding UE trajectory prediction model according to some embodiments of the disclosure;

[0048] Figure 9 is a block diagram of an NWDAF for performing a process for UE-assisted data collection for mobility prediction according to some embodiments of the disclosure;

[0049] Figure 10 and11 Shows the sharing of auxiliary information for training or retraining UE trajectory prediction between UEs according to some embodiments of the present disclosure;

[0050] Figures 12 to 14 Is a schematic block diagram of an exemplary embodiment of a network node according to some embodiments of the present disclosure; and

[0051] Figure 15 and 16 Is a schematic block diagram of an exemplary embodiment of a UE according to some embodiments of the present disclosure. Detailed Description

[0052] The embodiments set forth below represent information that enables those skilled in the art to implement the embodiments and show the best mode of implementing the embodiments. When reading the following description in view of the accompanying drawings, those skilled in the art will understand the concepts of the present disclosure and will realize applications of these concepts that are not specifically mentioned herein. It should be understood that these concepts and applications fall within the scope of the present disclosure.

[0053] Radio node: As used herein, a "radio node" is a radio access node or a wireless device.

[0054] Radio access node: As used herein, a "radio access node" or "radio network node" is any node in the radio access network (RAN) of a cellular communication network that operates to wirelessly transmit and / or receive signals. Some examples of radio access nodes include, but are not limited to, base stations (e.g., a new radio (NR) base station (gNB) in a 3rd Generation Partnership Project (3GPP) 5th Generation (5G) NR network or an enhanced or evolved Node B (eNB) in a 3GPP Long Term Evolution (LTE) network), high-power or macro base stations, low-power base stations (e.g., micro base stations, pico base stations, home eNBs, or the like), and relay nodes. A base station is also referred to herein as a radio base station (RBS).

[0055] Core network node: As used herein, a "core network node" is any type of node in a core network or any node that implements core network functions. Some examples of core network nodes include, for example, a Mobility Management Entity (MME), a Packet Data Network Gateway (P-GW), a Service Capability Exposure Function (SCEF), a Home Subscriber Server (HSS), or the like. Some other examples of core network nodes include nodes that implement an Access Management Function (AMF), a User Plane Function (UPF), a Session Management Function (SMF), an Authentication Server Function (AUSF), a Network Slice Selection Function (NSSF), a Network Exposure Function (NEF), a Network Function (NF) Repository Function (NRF), a Policy Control Function (PCF), a Unified Data Management (UDM), or the like.

[0056] Wireless device: As used herein, a "wireless device" is any type of device that accesses a cellular communication network (i.e., is served by a cellular communication network) by wirelessly transmitting and / or receiving signals to / from one radio access node (or multiple radio access nodes). Some examples of wireless devices include, but are not limited to, Machine Type Communication (MTC) devices and User Equipment devices (UE) in a 3GPP network.

[0057] Network node: As used herein, a "network node" is any node that is part of the RAN or the core network of a cellular communication network / system.

[0058] Note that the description given herein focuses on 3GPP cellular communication systems, and thus, 3GPP terms or terms similar to 3GPP terms are often used. However, the concepts disclosed herein are not limited to 3GPP systems.

[0059] Note that in the description herein, the term "cell" may be mentioned; however, in particular for 5G NR concepts, beams may be used instead of cells, and thus, it is important to note that the concepts described herein apply equally to both cells and beams.

[0060] As discussed above, for the UE trajectory prediction model of the Network Data Analytics Function (NWDAF) model to perform well, it is important to receive sufficient events regarding RBS transitions. This contradicts the current design of 4G / 5G networks where mobility handling signaling is kept to a minimum. Under the current design, when a UE is in the idle mode and moves within its tracking area list, there are no RBS handover events. The only events initiated by an idle UE are periodic Tracking Area Updates (TAUs) / Registration Area Updates (RAUs), which are done once per hour in the default configuration. This means that the NWDAF will not receive all the RBS transitions of an idle UE. This in turn will lead to poor performance of the NWDAF trajectory model in predicting the next RBS of the UE.

[0061] There can be various reasons for the performance of the NWDAF trajectory model not being good enough. The first scenario is that the performance has never been good, for example, because there has simply not been enough training data to achieve good performance. The second scenario is that the performance is good but deteriorates over time. The deteriorating performance can have several reasons, including: infrastructure changes, e.g., new roads have been opened / roads have been closed since the initial training of the model; network changes, e.g., RBSs have been added or removed since the initial training of the model; and holiday periods that make people move in a different way than before. It is worth noting that the performance problem may be specific to a certain area, for certain times, or even for UEs with a certain capability.

[0062] Certain aspects of the present disclosure and their embodiments can provide solutions to the foregoing or other challenges. According to one aspect, the present disclosure provides methods and systems for enabling a UE to participate in the training of a mobility prediction model. In some embodiments, this includes: downloading a trained model to the UE, causing the UE to compare the predictions from the model with its actual movement, and causing the UE to take action when the prediction is incorrect. Possible actions include, but are not limited to: (1) the UE provides the correct movement to the network so that the network can retrain; 2) the UE retrains the model locally with the correct inputs and uploads the obtained model to the network.

[0063] According to another aspect, the present disclosure provides methods and systems by which the system itself detects that the performance of the NWDAF's trajectory prediction service is not good enough (e.g., the associated Key Performance Indicator (KPI) fails to meet a predefined or preconfigured threshold), and the network takes measures to increase the number of base station transition events sent to the NWDAF. Thereby, the performance of the prediction service will increase.

[0064] Some embodiments may provide one or more of the following technical advantages. One advantage of this solution is a mobility prediction service with better performance. Another advantage is that the service can be used by multiple use cases, including paging and UPF reselection.

[0065] Figure 2 FIG. 4 shows an example of a cellular communication system 200 in which embodiments of the present disclosure may be implemented. In the embodiments described herein, the cellular communication system 200 may be a 5G system (5GS) including an NR RAN or an evolved packet system (EPS) including an LTE RAN. In this example, the RAN includes base stations 202-1 and 202-2, which are referred to as eNBs in LTE and gNBs in 5G NR, and control corresponding (macro) cells 204-1 and 204-2. Base stations 202-1 and 202-2 are generally referred to collectively herein as base station 202 and individually as base station 202. Similarly, (macro) cells 204-1 and 204-2 are generally referred to collectively herein as (macro) cell 204 and individually as (macro) cell 204. The RAN may also include a number of low-power nodes 206-1 to 206-4, which control corresponding small cells 208-1 to 208-4. The low-power nodes 206-1 to 206-4 can be small base stations (such as pico or femto base stations) or remote radio heads (RRHs) or the like. It is noted that, although not shown, one or more of the small cells 208-1 to 208-4 may alternatively be provided by the base station 202. The low-power nodes 206-1 to 206-4 are generally referred to collectively herein as low-power node 206 and individually as low-power node 206. Similarly, the small cells 208-1 to 208-4 are generally referred to collectively herein as small cell 208 and individually as small cell 208. The cellular communication system 200 also includes a core network 210, which is referred to as a 5G core (5GC) in 5GS and an evolved packet core (EPC) in EPS. The base stations 202 (and optionally also the low-power nodes 206) are connected to the core network 210.

[0066] The base stations 202 and the low-power nodes 206 provide service to wireless devices 212-1 to 212-5 in the corresponding cells 204 and 208. The wireless devices 212-1 to 212-5 are generally referred to collectively herein as wireless device 212 and individually as wireless device 212. The wireless device 212 is sometimes also referred to as a UE in this document.

[0067] Figure 3A wireless communication system is shown representing a 5G network architecture composed of core NFs, where the interaction between any two NFs is represented by point-to-point reference points / interfaces. Figure 3 Can be regarded as Figure 2 A specific implementation of the cellular communication system 200.

[0068] Viewed from the access side, Figure 3 The 5G network architecture shown in [reference] includes a plurality of UEs 212 connected to the RAN 202 or access network (AN) and the AMF 300. Generally, the R(AN) 202 includes a base station, such as an eNB or gNB or the like. Viewed from the core network side, Figure 3 The 5GC NFs shown in [reference] include the NSSF 302, AUSF 304, UDM 306, AMF 300, SMF 308, PCF 310, and application function (AF) 312.

[0069] The reference points of the 5G network architecture are used to develop detailed call flows in specification standardization. The N1 reference point is defined to carry signaling between the UE 212 and the AMF 300. The reference points for connecting between the AN 202 and the AMF 300 and between the AN 202 and the UPF 314 are defined as N2 and N3 respectively. There is a reference point N11 between the AMF 300 and the SMF 308, which means that the SMF 308 is at least partially controlled by the AMF 300. N4 is used by the SMF 308 and the UPF 314, so that the UPF 314 can be configured using the control signals generated by the SMF 308, and the UPF 314 can report its status to the SMF 308. Separately, N9 is the reference point for connecting between different UPF 314s, and N14 is the reference point for connecting between different AMF 300s. N15 and N7 are defined because the PCF 310 applies policies to the AMF 300 and the SMF 308 respectively. The AMF 300 requires N12 to perform authentication of the UE 212. N8 and N10 are defined because the AMF 300 and the SMF 308 require subscription data of the UE 212.

[0070] The 5GC network aims to separate the user plane (UP) and the control plane (CP). The UP transports user traffic, while the CP transports signaling in the network. In Figure 3In it, the UPF 314 is in the UP, and all other NFs (i.e., the AMF 300, SMF 308, PCF 310, AF 312, NSSF 302, AUSF 304, and UDM 306) are in the CP. Separating the UP and CP ensures that each plane's resources are scaled independently. It also allows the UPF to be deployed in a distributed manner separately from the CP functions. In this architecture, the UPF can be deployed very close to the UE to shorten the round-trip time (RTT) between the UE and the data network for some applications that require low latency.

[0071] The core 5G network architecture consists of modular functions. For example, the AMF 300 and SMF 308 are independent functions in the CP. The separated AMF 300 and SMF 308 allow for independent evolution and scaling. Other CP functions (such as the PCF 310 and AUSF 304) can be separated, as Figure 3 shown. The modular function design enables the 5GC network to flexibly support various services.

[0072] Each NF directly interacts with another NF. It is possible to use intermediate functions to route messages from one NF to another NF. In the CP, a set of interactions between two NFs is defined as a service, so that its reuse is possible. This service can achieve support for modularity. The UP supports interactions between different UPFs, such as forwarding operations.

[0073] Figure 4 shows a 5G network architecture that uses service-based interfaces between NFs in the CP instead of the point-to-point reference points / interfaces used in the Figure 3 5G network architecture. However, the NFs described above correspond to the Figure 3 NFs shown in Figure 4 . One service (or multiple services) etc. provided by an NF to other authorized NFs can be opened to the authorized NFs through the service-based interface. In Figure 4 , the service-based interface is indicated by the letter "N" followed by the name of the NF. For example, Namf for the service-based interface of the AMF 300 and Nsmf for the service-based interface of the SMF 308, etc. Figure 4 The NEF 400, NRF 402, and NWDAF 404 in Figure 3 are not shown in the above discussion of Figure 3 . However, it should be clarified that all the NFs depicted in Figure 4 can interact with the Figure 3 NEF 400 and NRF 402 as needed, but are not explicitly indicated in

[0074] Figure 3 and4 Some properties of the NFs shown in FIG. may be described as follows. The AMF 300 provides UE-based authentication, authorization, mobility management, etc. Even a UE 212 using multiple access technologies is basically connected to a single AMF 300 because the AMF 300 is independent of the access technology. The SMF 308 is responsible for session management and assigns Internet Protocol (IP) addresses to the UE. It also selects and controls the UPF 314 for data delivery. If the UE 212 has multiple sessions, different SMFs 308 may be assigned to each session to manage them separately and may provide different functionality according to the session. The AF 312 provides information about packet flows to the PCF 310 responsible for policy control to support Quality of Service (QoS). Based on this information, the PCF 310 determines policies regarding mobility and session management to operate the AMF 300 and SMF 308 appropriately. The AUSF 304 supports the authentication function for the UE or similar devices and, therefore, stores the authentication data for the UE or similar devices, while the UDM 306 stores the subscription data of the UE 212. The data network (DN) (which is not part of the 5GC network) provides Internet access or carrier services and similar services.

[0075] The NF may be implemented as a network element on dedicated hardware, as a software instance running on dedicated hardware, or as a virtualized function instantiated on a suitable platform (e.g., cloud infrastructure).

[0076] The following description and associated figures describe non-limiting examples of UE-assisted data collection for mobility prediction.

[0077] In a first aspect of the present disclosure, UE-assisted data collection is provided. Figure 5A and 5B FIG. shows a process of UE-assisted data collection for mobility prediction according to some embodiments of the present disclosure. In Figure 5A and 5B In the embodiments shown in FIG., the UE 212 continuously provides information to the NWDAF 404, and the NWDAF 404 can use this information to train or retrain a trajectory prediction model. The steps labeled 100 to 122 are the same as the steps with the same numbers in Figure 1 FIG., and therefore, their descriptions will not be repeated here.

[0078] However, in Figure 5A and 5BIn the embodiment shown, NWDAF 404 takes additional actions: downloading a trajectory prediction model and one or more performance KPIs to UE 212 (step 500). The trajectory prediction model downloaded to UE 212 is the trajectory prediction model of UE 212. The model downloaded to UE 212 may have been trained. UE 212 installs and runs the model, and when UE 212 moves, it compares the predicted movement of the model with the actual movement made (step 502). In this way, UE 212 can deduce whether the prediction is correct.

[0079] In Figure 5A and 5B the embodiment shown, when the prediction is incorrect (or generally, when the KPI is not met), information about the actual movement can be sent to the network as a handover event, for example (step 504). Note that, as used herein, a "handover event" is an event related to the handover of a corresponding UE 212 from one RBS to another RBS. Each handover event includes at least a timestamp, a unique UE identifier, and an identifier of the new RBS. Even if additional information (such as the previous RBS and the time spent in the previous RBS) can also be deduced by NWDAF 404 from previously received events, the event may include such information. The events received by NWDAF 404 (step 504) can be used as inputs for training / retraining the model (step 506). In some embodiments, UE 212 may only send an error notification, for example, a notification in the case where there is a difference between the predicted handover and the actual handover. Alternatively, UE 212 may also send notifications of the following: other events; or other types of information that NWDAF 404 may find useful for training / retraining the model; or even information that NWDAF 404 may find useful for other purposes. In one embodiment, the trained / retrained UE trajectory prediction model obtained from step 506 is updated in UE 212 (step 508). In one embodiment, updating the UE trajectory prediction model in UE 212 includes downloading the trained or retrained UE trajectory prediction model to UE 212. In another embodiment, updating the UE trajectory prediction model in UE 212 includes downloading updated parameters used by the UE trajectory prediction model in UE 212.

[0080] In some embodiments, NWDAF 404 may still continue to receive handover events from AMF 300, such as in step 116. Alternatively, NWDAF 404 may receive information only directly from UE 212 (such as in step 504) and not from AMF 300. In these scenarios, optional step 116 may not occur.

[0081] A variant of the above solution is a solution in which, instead of in the NWDAF 404, the retraining is done in the UE 212. This makes it a federated learning solution. In Figure 6A and 6B this variant is described.

[0082] Figure 6A and 6B show a process for UE-assisted data collection for mobility prediction according to some embodiments of the present disclosure. In the embodiments shown in Figure 6A and 6B the UEs 212 continuously train or retrain copies of their trajectory prediction models, which copies may ultimately be uploaded to the NWDAF 404. The steps labeled 100 to 122 are the same as the steps of the same number in Figure 1 and, therefore, their description will not be repeated here.

[0083] In the embodiments shown in Figure 6A and 6B the NWDAF 404 takes additional actions: downloading the trajectory prediction model and a performance KPI (or multiple performance KPIs) to the UE 212 (step 600). The model downloaded to the UE 212 may have been trained. The UE 212 installs and runs the model, and when the UE 212 moves, it compares the predicted movement of the model with the actual movement made (step 602). In this way, the UE 212 can deduce whether the prediction is correct.

[0084] In Figure 6A and 6BIn the embodiments shown, when the prediction is incorrect (or generally, when a KPI is not met), a local copy of the trajectory prediction model within UE 212 is trained or retrained (step 604). Immediately or eventually, the retrained trajectory prediction model is uploaded to NWDAF 404 (step 606), and the retrained model is incorporated into the main or global model (step 608). In some embodiments, UE 212 may send only the updated trajectory prediction model. Alternatively, UE 212 may also send a notification of an error event, other event, or other types of information that NWDAF 404 may find useful for training / retraining the model, or even information that NWDAF 404 may find useful for other purposes. In these embodiments, the model maintained by NWDAF 404 may be retrained independently of the model maintained by UE 212 (e.g., also using data from other UEs), and when UE 212 uploads its updated trajectory prediction model, NWDAF 404 may align the model from UE 212 with the model maintained internally by NWDAF 404.

[0085] In one embodiment, the trained / retrained UE trajectory prediction model obtained from step 608 is updated in UE 212 (step 610). In one embodiment, updating the UE trajectory prediction model in UE 212 includes downloading the trained or retrained UE trajectory prediction model to UE 212. In another embodiment, updating the UE trajectory prediction model in UE 212 includes downloading updated parameters used by the UE trajectory prediction model in UE 212 to UE 212.

[0086] In some embodiments, NWDAF 404 may still continue to receive transition events from AMF 300, such as in step 116. Alternatively, NWDAF 404 may receive information only directly from UE 212 (e.g., as in step 606) and not from AMF 300. In these scenarios, optional step 116 may not occur.

[0087] In this variant, each UE 212 retrains its local model using input from its own observations (step 604). The retrained local model or a set of parameters (such as weights in a neural network) is uploaded to the network (step 606), where NWDAF 404 incorporates the local model into the global model (step 608).

[0088] In some embodiments, multiple local models can be combined in this way, optionally retrained (step 116) using additional inputs from the AMF 300, and the resulting model can be used to serve different service functions (steps 118, 120, and 122) and downloaded to the UE again (step 600).

[0089] Regarding UE behavior, the UE 212 tracks the most recent n transitions as these are used as inputs to the model, where n can be configured via the RAN broadcast channel, non-access stratum (NAS) signaling, or UE device management. The recorded transitions are used as described above.

[0090] To select between the first or second solutions above, the UE 212 is provided with rules. The rules are supplied to the UE 212 in different ways, including NAS signaling extended for this purpose. Alternative signaling paths can be via radio resource control (RRC) or via the application layer.

[0091] When the network commands the UE 212 to observe mobility events, the UE 212 starts tracking its movement. In addition to replying to the RAN's reporting requests, the UE 212 also makes observations between these reports. Whether the UE 212 is in the CONNECTED state or the IDLE state, the UE 212 observes, for example, which base station provides the strongest signal. In this way, the UE 212 builds a list of transitions that can be used for training.

[0092] Regarding which UEs to select for obtaining UE-assisted data for mobility prediction, the following general guidelines can be considered. Downloading / uploading, running, and retraining in the UE 212 introduce new costs in terms of energy consumption, computing resources, and signaling. Increased battery consumption in the UE 212 is an issue. Therefore, it is important to select the correct set of UEs 212. Not all UEs 212 need to run the local model, and UEs 212 that run the local model may not always do so. Exemplary guidelines for selecting the UEs 212 to be involved include:

[0093] · Pick UEs 212 in a certain area / region where we see poor performance.

[0094] · Classify the UEs 212 (such as rarely moving, fast moving, etc.) and pick UEs 212 from the categories where the performance does not meet the standard.

[0095] · Pick UEs 212 without a battery or with a large battery; for example, vehicle-mounted UEs.

[0096] · Observe that UEs 212 of a certain category (e.g., UEs supporting a certain set of radio access technologies (RATs) and frequency bands) perform poorly; the selected UEs 212 will be from this category.

[0097] In addition, observing that the performance is poor during certain time periods of a day / week / year, the NWDAF 404 can assist the OAM in selecting a suitable time to report more data from the selected set of UEs 212.

[0098] In a second aspect, gap analysis is provided. A more advanced way for UE selection is to have the NWDAF 404 or the OAM perform an analysis to determine where there are gaps in the data. The underlying assumption is that gaps cause poor prediction performance. This is shown in Figure 7A and 7B during the process.

[0099] Figure 7A and 7B show a process according to some embodiments of the present disclosure, by which the system can detect that the performance of the trajectory prediction service of the NWDAF is not good enough and take measures to increase the number of base station handover events sent to the NWDAF 404. The steps labeled 100 to 122 are the same as the steps with the same numbers in Figure 1 and thus their description will not be repeated here.

[0100] However, in the embodiments shown in Figure 7A and 7B each prediction regarding the next RBS that the UE will eventually reach will have a certain accuracy. By comparing the prediction with the information in the handover events from the AMF 300, the prediction accuracy can be measured. A typical KPI is the percentage of correctly predicted handovers. This KPI will change over time. When the model is still untrained, the performance will be low. But for various reasons, the KPI of the trained model can also change over time. For example, when too little retraining data is provided. In our use case, when too many UEs generate too few events over too long a time period, too little data is provided.

[0101] When the performance of the prediction service is higher than a given KPI (or given KPIs), we define the prediction service as "ready". This KPI is provided by the OAM (step 100), or by the subscribers of the service (step 106), or by both. In the latter case, the OAM can provide the minimum KPI. When the service becomes "ready" or "not ready", the subscribers are notified (steps 104 and 108).

[0102] In Figure 7A and 7BIn the embodiment shown, the system determines whether the performance of the model is below a given KPI, e.g., whether the trajectory prediction service KPI is not met (step 700). If the performance of the model is acceptable, e.g., meets the KPI, NWDAF 404 notifies a subscriber (or subscribers) that the trajectory prediction service is ready and provides the trajectory prediction (step 702). If the performance of the model is unacceptable, NWDAF 404 notifies a subscriber (or subscribers) that the trajectory prediction service is not ready (704), and decides to take some action to improve the model performance (step 706). In Figure 7A and 7B In the embodiment shown, e.g., NWDAF 404 determines that it should notify the OAM that the KPI cannot be met (step 708), and the OAM can decide to take action to increase the data density (step 710).

[0103] In the above flow chart, NWDAF 404 is the entity that measures the performance of the model. It is NWDAF 404 that notifies the OAM, and the OAM is the entity that initiates the action. Many variations of the division of these tasks are possible:

[0104] · NWDAF 404 can directly contact the AMF 300 when the performance drops below a given threshold and request the AMF 300 to generate more events.

[0105] · The SF can perform its own performance measurement and report to NWDAF 404 or the OAM.

[0106] · NWDAF 404 can decide the action to be taken (instead of the OAM) and request the OAM to execute the action. The OAM can forward the request to the AMF 300 in such a case.

[0107] Actions that the OAM can take include but are not limited to commanding NWDAF 404 to download the local UE trajectory prediction model to the UE 212, as described in relation to the first aspect of the present disclosure. Other actions include:

[0108] Preventing the UE from becoming idle. One way for the OAM to request the AMF 300 to generate more events is to prevent a small number of UEs 212 from becoming idle. Thus, in one embodiment, the OAM sends a command to the AMF 300 to command the AMF 300 to keep the UE 212 connected. When the UE 212 does not enter the idle mode, it will generate events when performing an RBS handover. The AMF 300 then forwards these events to NWDAF 404 as an RBS transition.

[0109] If the OAM sends such a request and the UE 212 is idle, the AMF 300 can page the UE 212 according to the existing paging procedure (see Section 4.2.3.3 of 3GPP Technical Specification (TS) 23.502). Typically, after a certain period of inactivity, the UE212 will become idle. It is the RBS (e.g., gNB or eNB) that maintains this timer and initiates the process of becoming idle (see Section 4.2.6 of 3GPP TS 23.502). In one embodiment, the AMF 300 may set this timer to a value provided by the OAM; that is, the UE 212 needs to provide a time period for handover events even when there is no traffic going to or from the UE 212. This parameter can then be provided from the AMF 300 to the RBS. Note that this process can be per UE, so the "time period" parameter can be different for different UEs.

[0110] In an alternative embodiment, the OAM can communicate with the UE 212 via normal UP signaling. The UE 212 may have a special application, or the OAM may send a ping. In this case, paging will be initiated due to the first downlink UP packet from the OAM (all according to the existing procedure in Section 4.2.3.3 of 3GPP TS23.502). To prevent the UE 212 from becoming idle, the OAM can periodically re - send the downlink UP packet.

[0111] Smaller tracking area. Another way to obtain more mobility events from the UE would be by adjusting the tracking area list. A small group of UEs can receive a tracking area list spanning a smaller area. In this way, these UEs are forced to generate more TAU events as they move.

[0112] Shorter periodic update timer. When the UE is in the idle mode, it periodically performs a registration update (Section 5.3.2 of 3GPPTS23.501) to notify the network that it still exists. This results in an AMF event and includes the current RBS of the UE 212. The AMF 300 can forward this information as a transition to the NWDAF 404. The periodic registration update timer is part of the UE profile and is provided to the UE at registration. The UE configuration update procedure (Section 4.2.4.2 of 3GPP TS 23.502) can be used to provide a new value.

[0113] To request more mobility events from the OAM, it can command the AMF 300 to send the periodic registration update timer to the UE 212, where the timer has a small value (e.g., a few minutes instead of the default value of one hour).

[0114] User location service. Another way to feed additional transition events to the NWDAF 404 is to use the 3GPP user location service as described in Section 6.1.2 of 3GPP TS 23.273. In this scenario, the NWDAF 404 acts as an external client that periodically requests the location of the UE. In our use case, the format of the location data provided can be the RBS identifier.

[0115] In the above actions, existing processes are used to generate more events. Another way would be to introduce a new mechanism for obtaining UE location information.

[0116] New information element (IE). In one embodiment, a new IE can be added to the system information broadcast by each RBS (see 3GPP TS 36.300). If there are gaps in the dataset in a given area, the OAM commands the RBSs in that area to indicate in the system information that the UE 212 needs to provide location information. The UE 212 can do this using existing processes (e.g., sending periodic updates) or through new processes (e.g., providing location information to the AMF 300 via the NAS, where the AMF 300 forwards the location information to the OAM).

[0117] External metadata. Another way to provide more training data to the model would be to generate transitions of imaginary UEs from external metadata. One possible scenario would be to obtain the physical locations of the RBSs and place them on a map. The map can also store information about the roads, the types of roads, and the speed limits of the roads. Given this information, the imaginary UEs can perform trajectories along these roads. Then, such trajectories can be the basis for generating RBS transitions that are used to train the model in the NWDAF 404.

[0118] Non-mobile UEs. A special category of UEs 212 are those for which the NWDAF 404 has not received their transitions for a long time. This includes non-mobile UEs or UEs that only move within their tracking area list. Since the NWDAF 404 never receives training data for such UEs, it is difficult to perform trajectory prediction. One way to accommodate this category of UEs would be to have the AMF 300 periodically generate transitions in which the source RBS and the target RBS are equal.

[0119] Figure 8 A process for identifying a target UE 212 according to some embodiments of the present disclosure is shown. It can be before step 500 in the processes of Figure 5A and 5B or in Figure 6A and 6BThis process is performed before step 600 in the process of identifying such a process for the target UE 212, so as to select the UE 212 to which the UE trajectory prediction model is to be downloaded, in order to obtain auxiliary information for training / retraining the model. As shown, the process includes identifying the target UE 212 (step 800) from which auxiliary information for training / retraining the corresponding UE trajectory prediction model is to be obtained. In one embodiment, identifying the target UE 212 includes receiving an RBS handover event from at least one UE 212 (step 800A), and based on the received RBS handover event, identifying the target UE 212 from which auxiliary information for training / retraining the corresponding UE trajectory prediction model is to be obtained (step 800B).

[0120] In one embodiment, in order to identify the target UE 212 in step 800B, the process includes determining that the RBS handover event of the first UE among the received RBS handover events involves a handover from the first RBS to a second RBS that is not a neighbor of the first RBS (step 800B1), and identifying the first UE as the target UE 212 (step 800B2).

[0121] In another embodiment, in order to identify the target UE 212 in step 800B, the process includes: determining that the RBS handover event of each UE among the received RBS handover events involves a handover from the first RBS to a second RBS that is not a neighbor of the first RBS (step 800B3); identifying the plurality of UEs as potential target UEs for downloading the corresponding UE trajectory prediction model (step 800B4); and selecting a subset of the potential target UEs as one target UE (or multiple target UEs) for downloading the corresponding UE trajectory prediction model, where the subset of the potential target UEs includes the identified target UE (step 800B5). In one embodiment, the selection of the subset of the potential target UEs is completed based on one or more parameters (e.g., the presence of UEs in an area or region where the performance of the UE trajectory prediction model is below a threshold, UE category or type, UE speed, type of UE movement, UE behavior, RAT or frequency band supported by the UE, and / or UE battery capacity).

[0122] In an alternative embodiment, identifying the target UE in step 800 includes receiving the identification of the target UE from the OAM node.

[0123] Figure 9 is a block diagram of the NWDAF for performing UE-assisted data collection for mobility prediction according to some embodiments of the present disclosure. In Figure 9 the embodiment shown, the NWDAF 900 (corresponding to the NWDAF 404) includes a trajectory prediction module 902, a gap analysis module 904, and a filtering module 906.

[0124] The trajectory prediction module 902 receives information about the most recent N RBS transitions of a UE (referred to as "UE A"). The trajectory prediction module 902 predicts the next RBS of UE A.

[0125] In some embodiments, the trajectory prediction model 902 uses a number of recent transactions as input; for example, one embodiment uses the most recent four transactions. The transaction information includes the current and previous RBSs and the type of event that caused the transaction. If the event type is a handover, the NWDAF 900 knows that the UE is in the connected mode because that is the way the handover process is designed. This means that: the current and previous RBSs seen in this handover event are physical neighbors. By observing all handover events over a longer period of time, the NWDAF 900 can build a map of the RBSs of all physical neighbors. Of course, the map of physical neighbors can also be obtained in other ways; for example, as external metadata.

[0126] When a transition with a non-handover event type (e.g., periodic registration update) arrives, the NWDAF 900 can use the neighbor map to infer whether the current and previous RBSs are physical neighbors. If the current and previous RBSs are not physical neighbors, the NWDAF 900 can infer that there is a gap in the data of this UE. In other words, the UE is idle or has been idle, and while idle, the UE has passed one or more RBSs on its trajectory within its tracking area list and no mobility events have been created for the one or more RBSs.

[0127] In some embodiments, the NWDAF 900 can be designed to have a gap analysis module 904. This module generates a list of UEs that have seen their gaps. Another module can analyze this list and filter out UEs that have many or large gaps. The UEs from this filtered list can then participate in the training process as described above. The gap analysis module 904 can receive information about the most recent N RBS transitions of UE A, compare the RBS transitions to see if they involve transitions between known RBS neighbor pairs (the gap analysis module 904 also receives information about known RBS neighbor pairs), and generate a list of UEs with data gaps. This information is provided to the filtering module 906.

[0128] The filtering module 906 filters the list of UEs with data gaps to create a list of UEs from which additional data should be collected (a process also referred to as "data densification"). The UEs on this list are candidate targets into which the UE-based trajectory prediction module and performance KPIs discussed in detail above are to be downloaded and installed.

[0129] The previous section describes how the network selects UEs that will participate in data collection and retraining of the model. In some embodiments, the network may also select a group of UEs (e.g., a group of UEs currently located in a geographical location with poor prediction), and command those UEs to form a collaboration group. Information regarding the movement or updated retrained model can be shared directly (even device-to-device) among these UEs. In this way, group-based learning is provided.

[0130] For example, considering Figure 5A and 5B the process, UE 212 may send the information of step 504 to another UE (or other UEs) in the same collaboration group, as shown in Figure 10 step 1000 of Figure 6A and 6B the process, UE 212 may send the information of step 606 to another UE (or other UEs) in the same collaboration group, as shown in Figure 11 step 1100 of

[0131] In Figure 1 , 5A , 5B, 6A, and 6B, NWDAF 404 and AMF 300 communicate directly. In an alternative embodiment, NWDAF 404 and AMF 300 may communicate via OAM. This solution will give OAM more control over aspects such as which UEs to select, when to involve these UEs in the process, and what KPIs each UE should be measured against.

[0132] The above examples are described in the context of the mobility prediction use case. However, the ideas and concepts are equally applicable to other use cases, such as UE communication patterns, which involve questions such as what traffic is generated and / or consumed by the UE, at what time, and where.

[0133] Figure 12is a schematic block diagram of a network node 1200 according to some embodiments of the present disclosure. The network node 1200 can be a radio access node or a core network node. The radio access node can be, for example, base station 202 or 206. As shown, the network node 1200 includes a control system 1202, which includes one or more processors 1204 (e.g., central processing unit (CPU), application specific integrated circuit (ASIC), field programmable gate array (FPGA), and / or the like), a memory 1206, and a network interface 1208. The one or more processors 1204 are also referred to herein as processing circuitry. Additionally, the network node 1200 can include one or more radio units 1210, each radio unit 1210 including one or more transmitters 1212 and one or more receivers 1214 coupled to one or more antennas 1216. The radio unit 1210 can be referred to as radio interface circuitry, or can be part of radio interface circuitry. In some embodiments, the (one or more) radio units 1210 are located external to the control system 1202 and are connected to the control system 1202 via, for example, a wired connection (e.g., an optical cable). However, in some other embodiments, the (one or more) radio units 1210 and potentially also the (one or more) antennas 1216 are integrated with the control system 1202. The one or more processors 1204 operate to provide one or more functions of the network node 1200 as described herein. In some embodiments, the (one or more) functions are implemented in software, which is stored, for example, in the memory 1206 and executed by the one or more processors 1204.

[0134] Figure 13 is a schematic block diagram showing a virtualized embodiment of a network node 1200 according to some embodiments of the present disclosure. This discussion applies equally to other types of network nodes. Additionally, other types of network nodes can have a similar virtualized architecture.

[0135] As used herein, a "virtualized" radio access node is an implementation of network node 1200, where at least a portion of the functionality of network node 1200 is implemented as one or more virtual components (e.g., via one or more virtual machines executing on one or more physical processing nodes in one or more networks). As shown, in this example, as described above, network node 1200 includes control system 1202 and one or more radio units 1210. Control system 1202 includes one or more processors 1204 (e.g., CPU, ASIC, FPGA, and / or the like), memory 1206, and network interface 1208. Each radio unit 1210 includes one or more transmitters 1212 and one or more receivers 1214 coupled to one or more antennas 1216. Control system 1202 is connected to the one or more radio units 1210 via, for example, an optical cable or the like. Control system 1202 is connected to one or more processing nodes 1300 via network interface 1208, and the one or more processing nodes 1300 are coupled to one or more networks 1302 or are included as part of one or more networks 1302. Each processing node 1300 includes one or more processors 1304 (e.g., CPU, ASIC, FPGA, and / or the like), memory 1306, and network interface 1308.

[0136] In this example, the functionality 1310 of network node 1200 described herein is implemented in one or more processing nodes 1300 or is distributed in any desired manner between control system 1202 and one or more processing nodes 1300. In some particular embodiments, some or all of the functionality 1310 of network node 1200 described herein is implemented as virtual components executed by one or more virtual machines implemented in one or more virtual environments hosted by one or more processing nodes 1300. As will be understood by those of ordinary skill in the art, additional signaling or communication between the one or more processing nodes 1300 and control system 1202 is used to perform at least some of the desired functionality 1310. It is noted that in some embodiments, control system 1202 may not be included, in which case the one or more radio units 1210 communicate directly with the one or more processing nodes 1300 via one or more suitable network interfaces.

[0137] In some embodiments, a computer program is provided, the computer program comprising instructions which, when the computer program is executed by at least one processor, cause the at least one processor to perform the functionality of one or more of the following nodes according to any of the embodiments described herein: a network node 1200 or a node (e.g., a processing node 1300) that implements one or more functions of the functionality 1310 of the network node 1200 in a virtual environment. In some embodiments, a carrier comprising the foregoing computer program product is provided. The carrier is one of the following: an electronic signal, an optical signal, a radio signal, or a computer-readable storage medium (e.g., a non-transitory computer-readable medium such as a memory).

[0138] Figure 14 is a schematic block diagram of a network node 1200 according to some other embodiments of the present disclosure. The network node 1200 includes one or more modules 1400, each module 1400 implemented in software. The (one or more) modules 1400 provide the functionality of the network node 1200 described herein. This discussion applies equally to Figure 13 the processing node 1300, where the module 1400 may be implemented in one of the processing nodes 1300 or distributed over multiple processing nodes 1300 and / or distributed over the (one or more) processing nodes 1300 and the control system 1202.

[0139] Figure 15 is a schematic block diagram of a UE 1500 according to some embodiments of the present disclosure. As shown, the UE 1500 includes one or more processors 1502 (e.g., a CPU, an ASIC, an FPGA, and / or the like), a memory 1504, and one or more transceivers 1506, each transceiver 1506 including one or more transmitters 1508 and one or more receivers 1510 coupled to one or more antennas 1512. As will be understood by those of ordinary skill in the art, the (one or more) transceivers 1506 include radio front-end circuitry connected to the (one or more) antennas 1512, the radio front-end circuitry being configured to condition signals transmitted between the (one or more) antennas 1512 and the (one or more) processors 1502. The processor 1502 is also referred to herein as a processing circuit. The transceiver 1506 is also referred to herein as a radio circuit. In some embodiments, the functionality of the UE 1500 described above may be implemented fully or partially in software, the software being stored, for example, in the memory 1504 and executed by the (one or more) processors 1502. Note that the UE 1500 may include Figure 15Additional components not shown, such as, for example, one or more user interface components (e.g., input / output interfaces including a display, buttons, touch screen, microphone, speaker(s), and / or the like and / or any other components for allowing information to be input into UE 1500 and / or allowing information to be output from UE 1500), a power source (e.g., a battery and associated power circuitry), etc.

[0140] In some embodiments, a computer program is provided that includes instructions that, when executed by at least one processor, cause the at least one processor to perform the functionality of UE 1500 according to any of the embodiments described herein. In some embodiments, a carrier including the foregoing computer program product is provided. The carrier is one of the following: an electronic signal, an optical signal, a radio signal, or a computer-readable storage medium (e.g., a non-transitory computer-readable medium such as a memory).

[0141] Figure 16 is a schematic block diagram of UE 1500 according to some other embodiments of the present disclosure. UE 1500 includes one or more modules 1600, each module 1600 implemented in software. The module(s) 1600 provide the functionality of UE 1500 described herein.

[0142] Any suitable steps, methods, features, functions, or benefits disclosed herein may be performed by one or more functional units or modules of one or more virtual devices. Each virtual device may include a number of such functional units. These functional units may be implemented via a processing circuit and other digital hardware, the processing circuit may include one or more microprocessors or microcontrollers, and the other digital hardware may include a digital signal processor (DSP), dedicated digital logic, and the like. The processing circuit may be configured to execute program code stored in a memory, which may include one or several types of memories such as read-only memory (ROM), random access memory (RAM), cache, flash memory devices, optical storage devices, etc. The program code stored in the memory includes program instructions for executing one or more telecommunication and / or data communication protocols and instructions for performing one or more of the techniques described herein. In some implementations, the processing circuit may be used to cause the corresponding functional unit to perform the corresponding function according to one or more embodiments of the present disclosure.

[0143] Although the processes in the figures may show a particular order of operations performed by certain embodiments of the present disclosure, it should be understood that this order is exemplary (e.g., alternative embodiments may perform operations in a different order, combine certain operations, overlap certain operations, etc.).

[0144] Some exemplary embodiments of the present disclosure are as follows:

[0145] Embodiment 1: A method for UE-assisted data collection for mobility prediction performed by a user equipment UE, the method comprising: receiving (500, 600) a UE trajectory prediction model for predicting the UE trajectory from a telecommunication core network node; executing (502, 602) the UE trajectory prediction model to generate a predicted trajectory of the UE; comparing (502, 602) the actual UE trajectory with the predicted UE trajectory; and sending (504, 606) the result of the comparison between the actual UE trajectory and the predicted UE trajectory to the telecommunication core network node.

[0146] Embodiment 2: The method according to Embodiment 1, wherein sending the result of the comparison between the actual UE trajectory and the predicted UE trajectory comprises: sending (504) a transition event or a trajectory event.

[0147] Embodiment 3: The method according to Embodiment 1 or 2, wherein sending (504) the result of the comparison between the actual UE trajectory and the predicted UE trajectory comprises: sending error data describing the difference between the actual UE trajectory and the predicted UE trajectory.

[0148] Embodiment 4: The method according to Embodiment 2 or 3, further comprising: sending the result of the comparison between the actual UE trajectory and the predicted UE trajectory to another UE.

[0149] Embodiment 5: The method according to Embodiment 1, further comprising: training or retraining (604) the UE trajectory model after comparing (602) the actual UE trajectory with the predicted UE trajectory.

[0150] Embodiment 6: The method according to Embodiment 5, wherein sending the result of the comparison between the actual UE trajectory and the predicted UE trajectory comprises: uploading (606) the trained or retrained UE trajectory model.

[0151] Embodiment 7: The method according to Embodiment 5 or 6, further comprising: sending the trained or retrained UE trajectory model to another UE.

[0152] Embodiment 8: The method according to any one of Embodiments 1-7, wherein receiving (500, 600) the UE trajectory model from a telecommunication network node comprises: receiving the UE trajectory model from a Network Data Analytics Function NWDAF, and wherein sending (504, 606) the result of the comparison to the telecommunication core network node comprises: sending the result of the comparison to the NWDAF.

[0153] Example 9: A method for user equipment UE-assisted data collection for mobility prediction performed by a network node, the method comprising: identifying a target UE; downloading (500, 600) a UE trajectory prediction model for predicting the trajectory of the UE to the target UE; receiving (504, 606) from the target UE information generated by the UE trajectory prediction model; and using (506, 608) the received trajectory information to train or retrain a UE trajectory prediction model stored in and / or used by the network node.

[0154] Example 10: The method according to Example 9, further comprising: updating the UE trajectory prediction model in the target UE.

[0155] Example 11: The method according to Example 10, wherein updating the UE trajectory prediction model in the target UE comprises: downloading the trained or retrained UE trajectory prediction model to the target UE.

[0156] Example 12: The method according to Example 10, wherein updating the UE trajectory prediction model in the target UE comprises: downloading updated parameters used by the UE trajectory prediction model in the target UE to the target UE.

[0157] Example 13: The method according to any one of Examples 9-12, wherein identifying the target UE comprises: receiving a plurality of radio base station RBS handover events associated with at least one UE; determining that the RBS handover event of a first UE involves a handover from a first RBS to a second RBS that is not a neighboring pair with the first RBS; and identifying the first UE as the target for downloading the UE trajectory prediction model.

[0158] Example 14: The method according to Example 13, wherein identifying the target UE comprises: determining that the RBS handover event of each UE in the plurality of UEs involves a handover from a first RBS to a second RBS that is not a neighboring pair with the first RBS; identifying the plurality of UEs as potential targets for downloading the UE trajectory prediction model; and selecting a subset of the UEs as the target for downloading the UE trajectory prediction model from the plurality of UEs as potential targets, the subset of the UEs including the identified target UE.

[0159] Example 15: The method according to Example 14, wherein the subset of the UEs is selected according to a selection algorithm.

[0160] Example 16: The method according to Example 15, wherein the selection algorithm selects the subset of the UEs based on: the presence of UEs in areas or regions where the performance is below a threshold, UE category or type, UE speed or type of movement, UE behavior, supported UE radio access technology RAT or frequency band, and / or UE battery capacity.

[0161] Example 17: The method according to any one of Examples 9 - 16, wherein the network node includes a Network Data Analytics Function NWDAF or an Operation, Administration, and Maintenance OAM node.

[0162] Example 18: The method according to any one of Examples 9 - 12, wherein identifying the target UE includes: receiving an identifier of the target UE from an Operation, Administration, and Management OAM node.

[0163] Example 19: A method for User Equipment UE - assisted data collection for mobility prediction performed by a network node, the method including: receiving a plurality of Radio Base Station RBS handover events associated with at least one UE; determining that an RBS handover event of a first UE involves a handover from a first RBS to a second RBS that is not a neighboring pair with the first RBS; and identifying the first UE as a target for downloading a UE trajectory prediction model.

[0164] Example 20: The method according to Example 19, wherein identifying the target UE includes: determining that an RBS handover event of each UE among the plurality of UEs involves a handover from a first RBS to a second RBS that is not a neighboring pair with the first RBS; identifying the plurality of UEs as potential targets for downloading a UE trajectory prediction model; and selecting a subset of UEs from the plurality of UEs as potential targets as the target for downloading a UE trajectory prediction model, the subset of UEs including the identified target UE.

[0165] Example 21: The method according to Example 20, wherein the subset of UEs is selected according to a selection algorithm.

[0166] Example 22: The method according to Example 21, wherein the selection algorithm selects the subset of UEs based on: the presence of UEs in an area or region where the performance is below a threshold, UE category or type, UE speed or type of movement, UE behavior, supported UE Radio Access Technology RAT or frequency band, and / or UE battery capacity.

[0167] Example 23: The method according to any one of Examples 19 - 22, wherein the network node includes a Network Data Analytics Function NWDAF or an Operation, Administration, and Maintenance OAM node.

[0168] Example 24: A User Equipment UE (1500) includes: a transceiver (1506); a processor (1502); and a memory (1504) storing instructions executable by the processor, whereby the UE is operable to perform the steps according to any one of Examples 1 to 8.

[0169] Example 25: A network node (1200) includes: a network interface (1208); a processor (1502); and a memory (1206) storing instructions executable by the processor, whereby the network node is operable to perform the steps as described in any one of Examples 9 to 23.

[0170] Example 26: The network node (1200) as described in Example 25 includes a network data analysis function NWDAF (900).

[0171] Example 27: The NWDAF (900) as described in Example 25 further includes: a trajectory prediction module (902), a gap analysis module (904), and a filtering module (906). The trajectory prediction module (902): receives a user equipment UE radio base station RBS transition and predicts the next RBS of the UE. The gap analysis module (904): receives a UE RBS transition; determines that the RBS transition is from a first RBS to a second RBS that is a neighbor other than the first RBS, and identifies a data gap based on this determination; and generates a list of UEs associated with the identified data gap. The filtering module (906): filters the list of UEs associated with the identified data gap and generates a list of UEs from which additional data should be collected.

[0172] Example 28: A communication system provides user equipment UE-assisted data collection for mobility prediction. The communication system includes: a network data analysis function NWDAF and a UE. The network data analysis function NWDAF: downloads a UE trajectory prediction model for predicting the trajectory of the UE to a target UE; receives information generated by the UE trajectory prediction model from the target UE; and uses the received trajectory information to train or retrain a UE trajectory prediction model stored in and / or used by the network node. The UE: receives a UE trajectory prediction model for predicting the UE trajectory from the NWDAF, executes the UE trajectory prediction model to generate a predicted trajectory of the UE; compares the actual UE trajectory with the predicted UE trajectory; and sends the result of the comparison between the actual UE trajectory and the predicted UE trajectory to the NWDAF.

[0173] Example 29: In the communication system as described in Example 28, the NWDAF selects the target UE.

[0174] Example 30: The communication system as described in Example 28 further includes an operations, administration, and management OAM node that selects the target UE.

[0175] At least some of the following abbreviations may be used in this disclosure. If there is an inconsistency between the abbreviations, the usage pattern above it should be preferred. If listed multiple times below, the first listing should be preferred over any subsequent listing(s).

[0176] ·3GPP 3rd Generation Partnership Project

[0177] ·4G Fourth Generation

[0178] ·5G Fifth Generation

[0179] ·5GC Fifth Generation Core

[0180] ·5GS Fifth Generation System

[0181] ·AF Application Function

[0182] ·AI Artificial Intelligence

[0183] ·AMF Access and Mobility Management Function

[0184] ·AN Access Network

[0185] ·ASIC Application Specific Integrated Circuit

[0186] ·AUSF Authentication Server Function

[0187] ·CP Control Plane

[0188] ·CPU Central Processing Unit

[0189] ·DN Data Network

[0190] ·DSP Digital Signal Processor

[0191] ·eNB evolved Node B

[0192] ·EPC Evolved Packet Core

[0193] ·EPS Evolved Packet System

[0194] ·FPGA Field Programmable Gate Array

[0195] ·gNB New Radio Base Station

[0196] ·HSS Home Subscriber Server

[0197] ·IE Information Element

[0198] ·IP Internet Protocol

[0199] ·KPI Key Performance Indicator

[0200] ·LTE Long Term Evolution

[0201] ·MME Mobility Management Entity

[0202] ·MTC Machine Type Communication

[0203] ·NAS Non-Access Stratum

[0204] ·NEF Network Exposure Function

[0205] ·NF Network Function

[0206] ·NR New Radio

[0207] ·NRF Network Function Repository Function

[0208] ·NSSF Network Slice Selection Function

[0209] ·NWDAF Network Data Analytics Function

[0210] ·OAM Operation, Administration and Maintenance

[0211] ·PCF Policy Control Function

[0212] ·P-GW Packet Data Network Gateway

[0213] ·QoS Quality of Service

[0214] ·RAM Random Access Memory

[0215] ·RAN Radio Access Network

[0216] ·RAU Registration Area Update

[0217] ·RAT Radio Access Technology

[0218] ·RBS Radio Base Station

[0219] ·ROM Read Only Memory

[0220] ·RRC Radio Resource Control

[0221] ·RRH Remote Radio Head

[0222] ·RTT Round Trip Time

[0223] ·SCEF Service Capability Exposure Function

[0224] ·SF Service Function

[0225] ·SMF Session Management Function

[0226] ·TA Tracking Area

[0227] ·TAC Tracking Area Code

[0228] ·TAI Tracking Area Identity

[0229] ·TAU Tracking Area Update

[0230] ·TS Technical Specification

[0231] ·UDM Unified Data Management

[0232] ·UE User Equipment

[0233] ·UP User Plane

[0234] ·UPF User Plane Function

[0235] Those skilled in the art will recognize improvements and modifications to the embodiments of the present disclosure. All such improvements and modifications are considered to fall within the scope of the concepts disclosed herein.

Claims

1. A method for UE-assisted data collection for mobility prediction performed by a user equipment UE (212), the method comprising: Receiving (500; 600) from a network node of a cellular communication system (200) a UE trajectory prediction model for predicting the UE trajectory; Executing (502; 602) the UE trajectory prediction model to generate a predicted trajectory of the UE (212); Comparing (502; 602) the actual UE trajectory with the predicted trajectory of the UE (212); and Sending (504; 606) to a network node of the cellular communication system (200) the result of the comparison between the actual UE trajectory and the predicted trajectory of the UE (212).

2. The method according to claim 1, wherein, Sending (504; 606) the result of the comparison between the actual UE trajectory and the predicted trajectory of the UE (212) includes: sending (504) a transition event or a trajectory event.

3. The method according to claim 1 or 2, wherein Sending (504; 606) the result of the comparison between the actual UE trajectory and the predicted trajectory of the UE (212) includes: sending (504) error data describing the difference between the actual UE trajectory and the predicted trajectory of the UE (212).

4. The method according to claim 2, further comprising: Sending (1000) the result of the comparison between the actual UE trajectory and the predicted trajectory of the UE (212) to another UE.

5. The method according to claim 1, further comprising: After comparing (602) the actual UE trajectory with the predicted trajectory of the UE (212), training or retraining (604) the UE trajectory prediction model.

6. The method according to claim 5, wherein, Sending (504; 606) the result of the comparison between the actual UE trajectory and the predicted trajectory of the UE (212) includes: uploading (606) the trained or retrained UE trajectory prediction model.

7. The method according to claim 5 or 6, further comprising: Sending (1100) the trained or retrained UE trajectory prediction model to another UE.

8. The method according to claim 1, wherein Receiving (500; 600) the UE trajectory prediction model includes: receiving the UE trajectory prediction model from a network data analytics function NWDAF.

9. The method according to claim 1, wherein, Receiving (500; 600) the UE trajectory prediction model includes: receiving the UE trajectory prediction model from a management data analytics function, a non-real-time intelligent controller, or a near-real-time intelligent controller.

10. The method according to claim 1, wherein Sending (504; 606) the result of the comparison includes: sending the result of the comparison to a network data analytics function NWDAF.

11. The method according to claim 1, wherein, Sending (504; 606) the result of the comparison includes: sending the result of the comparison to a management data analytics function, a non-real-time intelligent controller, or a near-real-time intelligent controller.

12. A user equipment UE (212) for UE-assisted data collection for mobility prediction, the UE (212) being adapted to: Receiving (500; 600) from a network node of a cellular communication system (200) a UE trajectory prediction model for predicting the UE trajectory; Executing (502; 602) the UE trajectory prediction model to generate a predicted trajectory of the UE (212); Comparing (502; 602) the actual UE trajectory with the predicted trajectory of the UE (212); and Send (504; 606) the result of the comparison of the actual UE trajectory with the predicted trajectory of the UE (212) to a network node of the cellular communication system (200).

13. The UE (212) according to claim 12, wherein, The UE (212) is also adapted to perform the method according to any one of claims 2 to 11.

14. A user equipment UE (212) for UE-assisted data collection for mobility prediction, the UE (212) comprising: One or more transmitters (1508); One or more receivers (1510); And Processing circuitry (1502), associated with the one or more transmitters (1508) and the one or more receivers (1510), the processing circuitry (1502) being configured to cause the UE (212) to: Receive (500; 600) a UE trajectory prediction model for predicting the UE trajectory from a network node of a cellular communication system (200); Execute (502; 602) the UE trajectory prediction model to generate a predicted trajectory of the UE (212); Compare (502; 602) the actual UE trajectory with the predicted trajectory of the UE (212); and Send (504; 606) the result of the comparison of the actual UE trajectory with the predicted trajectory of the UE (212) to a network node of the cellular communication system (200).

15. A method for UE-assisted data collection for mobility prediction performed by a cellular communication system (200), the method comprising: Download (500; 600) a UE trajectory prediction model for predicting the trajectory of a target UE (212) to the target UE (212); Receive (504; 606) information related to training or retraining the UE trajectory prediction model from the target UE (212); and Use (506; 608) the received information to train or retrain the UE trajectory prediction model.

16. The method according to claim 15, further comprising: Update (508; 610) the UE trajectory prediction model in the target UE (212).

17. The method according to claim 16, wherein, Updating (508; 610) the UE trajectory prediction model in the target UE (212) includes: downloading (508; 610) the trained or retrained UE trajectory prediction model to the target UE (212).

18. The method according to claim 16, wherein, Updating (508; 610) the UE trajectory prediction model in the target UE (212) includes: downloading (508; 610) updated parameters used by the UE trajectory prediction model in the target UE (212) to the target UE (212).

19. The method according to claim 15, further comprising: Before downloading (500; 600) the UE trajectory prediction model to the target UE (212), identify (800) the target UE (212).

20. The method according to claim 19, wherein Identifying (800) the target UE (212) includes: Receiving (800A) a plurality of radio base station RBS handover events associated with at least one UE; Determining (800B1) that the RBS handover event of the first UE involves a handover from a first RBS to a second RBS that is not a neighbor of the first RBS; and Identify the first UE (800B2) as the target for downloading the UE trajectory prediction model.

21. The method according to claim 19, wherein, Identifying (800) the target UE (212) includes: Determining (800B3) that the radio base station RBS transition event of each UE among the multiple UEs involves a transition from a first RBS to a second RBS that is not a neighbor of the first RBS; Identifying (800B4) the multiple UEs as potential target UEs for downloading the corresponding UE trajectory prediction models; and Selecting (800B5) a subset of the potential target UEs as the target UEs for downloading the corresponding UE trajectory prediction models, the subset of the potential target UEs including the identified target UE (212).

22. The method according to claim 21, wherein, Selecting (800B5) the subset of the potential target UEs includes: selecting the subset of the potential target UEs based on at least one parameter.

23. The method according to claim 22, wherein, The at least one parameter includes: the presence of UEs in an area or region where the performance of the UE trajectory prediction model is below a threshold, UE category or type, UE speed, type of UE movement, UE behavior, radio access technology RAT or frequency band supported by the UE, and / or UE battery capacity.

24. The method according to any one of claims 15 to 23, wherein, At least some of the steps of the method are performed by a network data analytics function NWDAF or an operation, administration, and maintenance OAM node.

25. The method according to claim 19, wherein Identifying (800) the target UE (212) includes: receiving the identification of the target UE (212) from an operation, administration, and management OAM node.

26. A system for user equipment UE-assisted data collection for mobility prediction, the system comprising: At least one network node for a cellular communication system (200), the at least one network node being adapted to: Download (500; 600) a UE trajectory prediction model for predicting the trajectory of a target UE (212) to the target UE (212); Receive (504; 606) information related to training or retraining the UE trajectory prediction model from the target UE (212); and Use (506; 608) the received information to train or retrain the UE trajectory prediction model.

27. The system according to claim 26, wherein, The at least one network node is further adapted to perform the method according to any one of claims 16 to 25.

28. The system according to claim 26 or 27, wherein, Each network node of the at least one network node includes a processing circuit (1204; 1304), and the processing circuit (1204; 1304) is configured to cause the network node to perform at least one of the download, receive, and use operations.

29. A method for user equipment UE-assisted data collection for mobility prediction performed by a network node, the method including: Receiving (800A) multiple radio base station RBS transition events associated with at least one UE; And Identifying (800B) a target UE from which to obtain information for training or retraining a corresponding UE trajectory prediction model based on the multiple RBS transition events.

30. The method according to claim 29, wherein, Identifying (800B) the target UE includes: Determining (800B1) from among the multiple RBS transition events that the RBS transition event of a first UE involves a transition from a first RBS to a second RBS that is not a neighbor of the first RBS; and Identify the first UE (800B2) as the target UE.

31. The method according to claim 29, wherein Identifying (800B) the target UE includes: Determining (800B3) for each UE among the plurality of RBS transition events that the RBS transition event of the UE involves a transition from a first RBS to a second RBS that is not a neighbor of the first RBS; Identifying (800B4) the plurality of UEs as potential target UEs for downloading the corresponding UE trajectory prediction model; and Selecting (800B5) a subset of the potential target UEs as the target UEs for downloading the corresponding UE trajectory prediction model, the subset of the potential target UEs including the identified target UE.

32. The method according to claim 31, wherein, The subset of the potential target UEs is selected based on one or more parameters.

33. The method according to claim 32, wherein, The one or more parameters include: the presence of UEs in an area or region where the performance of the UE trajectory prediction model is below a threshold, UE class or type, UE speed, type of UE movement, UE behavior, supported radio access technology RAT or frequency band, and / or UE battery capacity.

34. The method according to any one of claims 29 to 33, wherein The network node includes a network data analytics function NWDAF or an operation, administration, and maintenance OAM node.

35. A network node for user equipment UE-assisted data collection for mobility prediction, the network node being adapted to: Receive a plurality of radio base station RBS transition events associated with at least one UE; and Identify, based on the plurality of RBS transition events, a target UE from which to obtain information for training or retraining a corresponding UE trajectory prediction model.

36. The network node according to claim 35, wherein, The network node is further adapted to perform the method according to any one of claims 30 to 34.

37. A network node (1200) for user equipment UE-assisted data collection for mobility prediction, the network node (1200): A processing circuit (1204; 1304), configured to cause the network node (1200) to: Receive a plurality of radio base station RBS transition events associated with at least one UE; and Identify, based on the plurality of RBS transition events, a target UE from which to obtain information for training or retraining a corresponding UE trajectory prediction model.

Images