Methods for inter-frequency ai / ml load balancing

Abstract

Currently, L3 mobility relies on event-based measurement reports for handovers. By utilizing AI / ML algorithms, the UE can predict measurement events directly, providing the network with advance guidan

Description

The present disclosure relates to apparatus and methods to enable the artificial intelligence / machine learning, "AI / ML", enhancement for New Radio, "NR" Mobility in a radio access network, "RAN". In particular, it defines new procedure / protocol, and AI / ML model requirements in accordance with a 3rd Generation Partnership Project, "3GPP", standard of a wireless communication system. BACKGROUND 5G NR is the latest cellular generation. 5G NR features several new innovations that allow for higher throughput, lower latency and extreme flexibility. Some of these important innovations that are relevant to this invention: Beam-based procedures: The device will take in to account the beams in a cell in several procedures to better accommodate the advancements in multiple-input multiple-output, "MIMO" and beamforming seen over the last decade. Ultra-lean carriers: Reduction in the number of “always-on” signals where the network may broadcast reference signals a lot more infrequently compared to previous generations, and allow a network to reduce the amount of system information broadcasted. More efficient state transitions: A new radio resource control, "RRC", state is introduced, RRCJNACTIVE. In RRC_INACTIVE, the user equipment, "UE", performs similar actions as in RRCJDLE, i.e. measuring and performing the cell reselection procedure to ensure that the UE is camping on the best cell. The network will save the UE context in the gNodeB, "gNB", and the UE will save the RRC configuration. This ensures that the state transition from RRCJNACTIVE and RRC_CONNECTED can be completed in a much smaller number of steps compared to moving from RRCJDLE to RRC_CONNECTED. In a network where there are a lot of state transitions, this can reduce latency and improve capacity as there is a lot less need for control signals to occupy capacity and resources. 5G NR procedures Below some relevant 5G NR procedures are explained that are relevant to understanding of the invention. Random access and state transitions In 5G NR there are three different RRC states: RRCJDLE, RRCJNACTIVE and RRC_CONNECTED. To move to RRC_CONNECTED, a UE must first synchronize and connect to a cell, which it does through the random access procedure. In the 4-step random access procedure, which is the procedure introduced for the first release of 5G NR, the procedure would include a Msg1 and a Msg2. Msg1 consist of a preamble sent on the Random Access Channel, "RACH", which is signals a number from 1 to 64 identifying the UE. Msg2 is the Random Access Response which contains a timing advance to synchronize the UE, as well as an uplink grant to send Msg3. Msg3 contains an RRC message and Msg4 contains the reply to the first RRC message as well as a contention resolution Medium Access Control Element, "MAC CE", to resolve any contention. For 2-step random access the MsgA consists of both the preamble and the first RRC message. MsgB consist of the random access response to synchronize the UE, the reply to the first RRC message as well as contention resolution. To move to RRC_CONNECTED from RRCJDLE, the RRC Setup procedure is triggered. RRC Setup procedures establishes a Signalling Radio Bearer 1, "SRB1" connection along with basic radio configurations. This means that the RRC message RRCSetupRequest will be included in Msg3 and the RRC message RRCSetup may be included in Msg4. After the RRC Setup procedures the network may also have to acquire capabilities, establish Access Stratum, "AS", security before data can transmitted. To move to RRC_CONNECTED from RRCJNACTIVE, the RRC Resume procedure is triggered. These procedures allow for the re-establishment of the full RRC connection, as well as resuming the AS security. This means that the RRC message RRCResumeRequest will be included in Msg3 and the RRC message RRCResume may be included in Msg4. It should be noted that MAC random access procedures are often independent of the RRC procedures, which means that the random access procedures may in general be the same for RRC Setup, RRC Resume, RRC Re-establishment and RRC reconfiguration with sync. RRC Release One way to enter RRC idle or RRC inactive mode is by the network releasing the UE through the RRC release procedures. The RRC release procedures are initiated when the UE receives a RRCRelease message from the gNB. The RRCRelease message sent from the gNB may in turn have been triggered by the either the gNB or the Access and Mobility management Function, "AMF". This can for instance be due to any of the following reasons: • Load balancing • Re-direction (both in RRC idle and RRC inactive) to other frequencies or Radio Access Technologies "RATs". • UE context release triggered by the Core Network, "CN", of the AMF • Suspend indication to send the UE to RRC inactive • Failure to retrieve UE context when UE resumes RRC connection from RRC inactive System information System information is information that is broadcasted by a cell for a wide range of purposes. System information is divided into a set of System Information Blocks, "SIB". Some system information is required for a UE to access a cell. Without having acquired these system information blocks the UE may not be allowed to access a cell. An example of such a SIB is SIB1 which contains access information, for instance the Public Land Mobile Network, "PLMN" of the cell, the cell identity, the tracking area code as well as cell selection information. SIB1 also contains the serving cell radio configuration. Another set of SIBs contain information on other frequencies and RATs for the purpose of idle and inactive mode cell reselection as well as related parameters. These are for instance in SIB2-SIB5. Connected mode mobility 5G NR connected mode mobility functions similar to other cellular standards. A standard connected mode handover is performed only when triggered by the gNB. The full usual procedure is as follows (seen in Figure 1): 1) A UE is configured with measurement configuration via a RRCReconfiguration procedure, which includes a Measurement Report NR, "MeasReportNR" which instructs when a UE shall report measurements and also includes the Measurement Object NR "MeasObjectNR" which gives details on how to perform the measurement. The UE may also be configured with a Measurement Object for EUTRA that specifies a carrier frequency on which the UE should perform measurements, "MeasObjectEUTRA", which is configured to measure 4G Evolved Universal Terrestrial Radio Access, "E-UTRA", cells. 2) UE performs neighbour cell measurements according to the measurement configuration, which are configured by the network. 3a) A measurement report (of a neighbour cell) is triggered and, 3b) is sent to the gNB. 4) Inter-node procedures whereby the Source evolved NodeB, "eNB" sends a Handover Request to Target eNB after having decided whether to trigger a handover and the Target eNB send a Handover Request Acknowledge to Source eNB. 5) The Source eNB triggers a handover command which is sent to the UE. This handover command consists of the RRC message RRCConnectionReconfiguration, containing the mobilityControlInfo field. 6) UE prepares for handover and performs a handover via the random access procedure. A measurement configuration consists of a set of measurement objects and measurement reporting configuration for Radio Resource Management, "RRM" purposes. These measurement objects define the configurations of the measurements that a UE shall perform in an RRC connected state. A measurement object is related to an either a frequency or a RAT. There are for instance intra-RAT NR Measurement objects MeasObjectNR, which may configure intra or inter-frequency measurements. The MeasObjectNR object contains configurations such as the carrier frequency, measurement bandwidth, Synchronization Signal Block, "SSB" configurations, neighbour cell configurations, specific cells to measure, cells not to measure and report and many more configurations. There are also Inter-RAT measurement object that contains all of the necessary configurations to perform connected mode measurements of E-UTRAN in the form of MeasObjectEUTRA, UTRA in the form of MeasObjectUTRA, GSM Edge Radio Access Network GERAN in the form of MeasObjectGERAN, CDMA2000 in the form of MeasGbjectCDMA2000, or Wireless Local Area Network "WLAN" in the form of MeasObjectWLAN. The measurement reporting configuration contains rules that define when a measurement report containing measurements shall be sent. These are typically defined by measurement reporting triggering conditions, thresholds, and configuring what measure shall be reported. Measurement reporting triggering conditions are also called measurement events. Other configurations may also include how many cells shall be reported and whether any additional information shall be reported, such as whether WLAN, Bluetooth or sensor information shall be reported. The UE location may also be included in the measurement report. A measurement identity is configuration which links one or more measurement objects to one or more reporting configurations. A quantity configuration defines the measurement filtering used for event evaluation and the reported measures. A measurement gap is a gap defined by the network, where the UE may switch away from its other configured responsibilities, such as monitoring for downlink and uplink assignments, in order to measure according to the configuration set by the network. This is mostly used for interfrequency or inter-RAT measurements. The measurement report will contain the measurement identity, as well as serving cell measurements. It may also contain the neighbour cell measurement for the specific RAT that is configured along with other configured additional information as explained above. Connected mode measurement events In order to trigger the measurement reporting, there are a set of measurement events that give specific conditions for when a measurement report shall be sent. A set of such well-known examples are the A1 to A5 and Event B1 and B2 events. These are defined as following: Event A1 (Serving becomes better than a threshold) Event A2 (Serving becomes worse than a threshold) Event A3 (Neighbour becomes offset better than Serving) Event A4 (Neighbour becomes better than threshold) Event A5 (Special Cell, "SpCell", becomes worse than thresholdl and neighbour becomes better than threshold2) Event B1 (Inter-RAT neighbour becomes better than threshold) Event B2 (Primary Cell, "PCell", becomes worse than thresholdl and Inter-RAT neighbour becomes better than threshold2) The signal type that is measured can either be the Reference Signal Received Power, "RSRP", which is the absolute strength of a measured signal, or the Reference Signal Received Quality, "RSRQ", which is the relative strength of a measure signal compared to noise and interference. Event A3 can thus be configured to be “Neighbour cell RSRP becomes a configured offset better than Serving cell RSRP”. For an event, there will be an entering and leaving condition, and a time during which the entering or leaving condition shall be true (timeToTrigger) before the measurement report is triggered. This to prevent frequent triggered measurement reports. For instance, for event A1 the entering condition is “Ms - Hys >Thresh”, and the leaving condition is “Ms + Hys <Thresh”, where Ms is the measurement result of the serving cell, Hys is the configured hysteresis parameter, and Thresh is the configured threshold. 5G NR also defines a number of other measurement events. For instance, Event B1 triggers when an inter-RAT neighbour becomes better than a threshold, event B2 triggers when the PCell becomes worse than a first threshold and the inter-RAT neighbour becomes better than a second threshold. Some enhanced measurement events for instance depend on the measured interference, e.g. Event 11, and / or depend on UE distance from two reference locations e.g. Event D1, and / or depend on whether time measured is above a threshold, e.g. CondEvent T1. Idle and inactive mode operation Idle and inactive mode mobility is based on a UE autonomously performing measurements and deciding according to some rules whether a UE shall re-select to another cell or not to camp on. During cell selection, the UE identifies suitable cells, which is according to a cell suitability criteria based on signal strength and signal quality measurements. After identifying one or several suitable cells, the UE can choose any of them. For instance, the UE can select the cell with the strongest signal strength and signal quality within a PLMN. Cell selection can be performed following PLMN selection, which may for example be after a UE is turned on, or after being released by a network, or during the RRC re-establishment procedure, and / or a number of other cases. A cell may be classified into a number of different types of cells. A suitable cell is a cell that fulfils the cell selection criteria, is not barred etc., and is part of the tracking area of the UE -thus a normal type of cell. An acceptable cell is a cell on which the UE is only allowed to camp for specific reasons such as emergency cases, but the cell cannot be barred, and the cell selection criteria need to be fulfilled. A UE only camps on such a cell if it cannot find a suitable cell. A reserved cell is a reserved if the system information indicates that it is reserved. When camped on a cell, the UE shall perform the cell reselection procedures which includes searching and detecting cells and camping on a better or more suitable cell. In the cell reselection procedure, the UE searches intra-frequency cells, inter-frequencies cells and inter-RAT cells following the signalling by the serving cell. Each frequency, either inter-RAT or intra-RAT, may have a specific cell reselection priority. The cell reselection algorithm is designed to ensure that the UE chooses a cell with highest priority, given that it is not barred or not allowed to camp on. A UE shall always select an inter-frequency or inter-RAT cell with a higher priority over a lower priority cell. If frequencies of equal priority are detected, then the UE shall rank all of the cells, where the ranking metric is based on signal strength and signal quality measurements and then choose the best candidate. The UE then camps on the newly reselected cell. There are also certain rules on for how long a new cell shall be better than the serving cell before camping on the new cell. This parameter is called Treselection and can be specific for a RAT or for other cases. There are also thresholds for the signal strength of signal quality that may need to be fulfilled before selecting a new cell to camp on, which may depend on whether the cell is lower or higher priority, and depend on whether the cell is inter-frequency or inter-RAT. As part of the idle mode and inactive procedures, the UE also checks whether a cell is barred or not. If a cell is barred or not, the UE is not allowed to connect to the cell, not allowed to camp or consider the cell for cell reselection. In 4G, the barring bit is signalled in a System Information Block Type 1, "SIB1" while in 5G NR, the barring bit is signalled in a Master Information Block, "MIB". Inter-RAT may be considered any other than the current RAT. As an example for 4G E-UTRA UE, the following may be considered inter-RAT: 2G, 3G, 4G Narrow Band Internet-of-Things "NB-loT", 5G NRor6G. TECHNICAL PROBLEM The 3GPP, study on AI / ML for mobility in NR and on AI-ML based Inter-frequency measurement prediction set out objectives focusing on mobility enhancement in the RRC_CONNECTED mode over air interface, and predicting inter-frequency measurements to decrease configuring measurement gaps at the UE within the same network, respectively. These studies were published at the 3GPP TSG RAN Meeting #103, RP-240082, Maastricht, Netherlands, 18-21 March 2024, and 3GPP TSG-RAN WG2 Meeting #R2-127, R2-2407113, Maastricht, Netherlands, 19-23 August 2024. There is a significant amount of overhead / cost for enabling inter-frequency measurements for UEs which affect: • Data transmission. A UE may have to interrupt data transmissions inter-frequency measurements. This may also affect overall capacity of the network, since the UE is not performing any measurements, but is still connected to the network. • UEs battery energy consumption. Due to the uncertainty of the radio conditions a gNB will typically request the UEs to measure inter-frequency cells even when they are not in the inter-frequency coverage area, consuming additional energy and affecting the UEs battery life. As a result, a lot of reporting is required for each UE. The innovation aims to tackle the need to report massive amounts of information to support existing procedures for NR or future 6G mobility. SOLUTION TO PROBLEM In this invention, we are solving at least one problem related to the inter-frequency measurement reporting prediction procedure for roaming / Handover between two networks. By innovating the UE and RAN capabilities for performing AI / ML, new procedures for enhancing Mobility and reducing the required measurements / signalling for the procedures becomes possible. The invention can be used for enhancing the 3GPP mobility procedure when one public 5G / 6G network cell and another private 5G / 6G network cell are operating at different frequency bands. When the UEs associated with the public network are reporting the measurements, they can simultaneously use this measurement to predict the private network cell measurement using AI / ML models. FIG. 2 shows a high-level schematic representation of a system in which the invention for enhancing the mobility scenario may be employed. The description below may apply to evolved Node Bs "eNBs", gNBs, Next Generation Radio Access Networks "NG-RAN", or Next Generation evolved Noted Bs "NG-eNB". For instance, in some cases the below description may be for an eNB, but it may also apply for NG-eNB, as comparably an eNB is connected to an Evolved Packet Core "EPC", while NG-eNB is connected to a 5G Core network "5GC". In some cases below, there may also be methods explicitly for an NG-eNB, e-gNB, gNB, or NG-RAN. For understanding, similar concepts from 4G and 5G and potential equivalents in a 6G system are described: 4G and 5G RRC connected state - UE having established a connection with a RAN, i.e. a cell, gNB or similar identity. Cell - This may also be a different concept in a 6G system. For instance in a cell-less case a UE may attach, connect and be associated with a beam or other similar identity. RRC idle - UE not in a RRC connected state, i.e., not having established a connection. RRC idle also means that UE will be camping on a cell or similar identity and then performing measurements and evaluating to find a better cell or similar identity. 5G RRC inactive - UE in a state similar to RRC idle where the UE stores the RRC configuration and resumes the RRC connection using the configuration. The network also stores the UE context and uses it to restore the UE connection. 5G RRC procedures (RRC Setup, RRC Resume, RRC Re-establishment, RRC Reconfiguration) - Any procedures that aims to establish a connection with a cell, a gNB or similar identity. For instance a procedure that aims to establish a 5G-6G Dual Connectivity setting with a 5G and 6G cell. Random access - The process of synchronizing the MAC layer via sending a preamble and receiving a message that synchronizes the uplink, as well as following messages to resolve any contention. Radio Link Failure - Failure of the radio link, which may be a failure based on measured radio signals, or based on operation in the cell, such as a number of retransmissions, random access failures, the radio beams failing etc. After the radio link failure the UE may try to reselect to another cell and re-establish the RRC connection. Handover - Performing active mobility to another cell, gNB or similar identity. Releasing RRC connection - The UE is released via a messages such as RRC Release that releases the RRC connection the UE has to one or more cells. This may also include the UE. The present disclosure also provides a list of information elements "lEs", apparatus and method for configuring UE and gNB node by AI / ML models according to the aforementioned UE / gNB-RAN ID list, so that the particular UE would work according to the control of the Handover procedures in 3GPP. One example discussed below is the UE / gNB-RAN using A3 messages for inter-frequency measurements. FIG. 3 shows a typical inter-frequency handover scenario, by way of background explanation. In FIG 3, when serving cell RSRP is lower than a threshold, an A2 event may be is triggered and UE send a measurement report to gNB. The gNB typically decides to activate inter-frequency measurements and configures the UE to start measuring inter-frequency neighbours. In this case, due to the need for the UE circuitry to switch frequency, the data transmission may be interrupted as there may need to be measurement gaps configured to the UE. UEs are typically asked to measure inter-frequency cells even when they are not in the interfrequency coverage area, consuming additional energy, affecting battery life. If the UE is in inter-frequency coverage area, a UE will search, detect and measure cells in the target cell frequency and an A3 inter-frequency measurement report may be sent to the gNB that will then trigger the handover, subject to the gNB handover algorithm. Other types of measurement triggers may also be configured. For instance, a UE may be configured to periodically report the inter-frequency measurement results. SUMMARY OF INVENTION This invention is divided up into different parts: 1. Configuring inter-frequency predictions; 2. Performing inter-frequency predictions; 3. Reporting inter-frequency predictions. Configuring inter-frequency predictions In one aspect of the invention, the UE is configured to predict the coverage condition for an interfrequency neighbour cell based on measurements of an intra-frequency cell, or another inter frequency cell. For instance, if the UE is configured to measure intra-frequency cells, the UE may be configured to predict the coverage of an inter-frequency cell. In another example, the UE may use measurements of a first inter-frequency cell to predict the coverage of a second inter-frequency cell. In another aspect of the invention, the UE may be configured to predict the coverage condition for an inter-frequency neighbour cell based on measurements performed of one or more intra and inter-frequency cells. For instance, the UE may be configured to measure one intra-frequency and a first inter-frequency cell and then using this the UE may predict the coverage of a second inter-frequency cell. This can for instance be useful in the case where measuring of a first inter-frequency cell may not require measurement gaps, or where the first inter-frequency cell may not require significant power consumption. For instance, the intra-frequency and the first inter-frequency cell may be in Frequency Range 1, "FR1", while the predicted cell or frequency may be in Frequency Range 2, "FR2". In this way, data transmissions in the network can be reduced and UE battery energy consumption reduced by prediction, rather than measurement of inter-frequency coverage conditions. In one aspect of the invention, an intra-frequency cell is configured with the option for the UE to predict inter-frequency coverage of a specific frequency. In other words, that a UE may be required to perform predictions on another frequency, based on measurements on a cell that the UE has been configured to measure, for instance an intra-frequency cell. To configure whether a UE shall perform predictions of an inter-frequency cell, there may be the following options: An intra-frequency cell is configured with the option to perform inter-frequency predictions of one or more inter-frequencies. o This can be useful in the case where not all cell-sites have multiple frequencies cells. For instance, for some cells, the UE may not be configured to provide any inter-frequency predictions. An intra-frequency cell is configured with the option that there may be inter-frequency predictions, and the specific frequencies where to perform predictions may be configured elsewhere, for instance it may be a general measurement configuration that applies to all cells. In one aspect of the invention, the predicted inter-frequency cell can be in a different Frequency Range, "FR". This means that the inter-frequency cell may for instance be in FR2, where the intra-frequency cell is FR1. In another aspect of the invention, the UE is only able to perform predictions and / or send or report a prediction, such as to send or report a prediction to a network, of an inter-frequency cell or inter-frequency under certain conditions. Example of this include: The inter-frequency or inter-frequency cell is within a certain frequency range away from an intra-frequency cell. If the inter-frequency or inter-frequency cell is in a certain frequency range o A UE measuring an intra-frequency in FR1 may not be allowed to provide predictions for FR2. The UE may only be allowed to perform prediction if a specific cell or frequency has been configured to perform predictions. o A specific inter-frequency cell can be configured specifically to not allow for predictions. This can also mean that the inter-frequency cell is configured to allowed for predictions. o A specific intra-frequency cell can be configured to not allow for inter-frequency predictions. This can also mean that an inter-frequency cell is configured to allowed for interfrequency predictions to be performed. o There may for instance be provided a predictionAllowedList or predictionNotAllowedList for this purpose. If an inter-frequency prediction is considered reliable enough. o For instance, if the reliability of prediction is below some type of threshold, then the UE may not provide a prediction to a network, else it does provide a prediction. The cell is configured as part of a measurement object. o If a cell is detected on a specific frequency, for instance a measured inter or intra-frequency, and it has not been configured as part of the measurement object, the UE may not be allowed to perform an inter-frequency prediction for this cell. This is important as a measurement object may specify specific cells to measure, but the UE may also detect other cells that have not been configured. o This may be configured in a cellsToAddModList, of the 5G NR measurement object, i.e. a list of cells to add or modify. o The intra-frequency cell shall also not be a part of any excluded cells list. o There may be a list of cells configured where the UE is allowed to perform inter frequency predictions. The specific inter-frequency cell that is being predicted has not been measured. The specific inter-frequency cell has not been measured recently. o For instance if the measurement gaps do not allow for continuous measurements, the prediction may be able to fill in gaps. A UE may be configured with a threshold for an intra-frequency cell, where if the signal strength or signal quality of the intra-frequency cell is below a threshold, the UE does not attempt to perform a prediction or does not report any predictions for one or more inter-frequency cells. This can be configured per cell, per frequency or it may be hardcoded. For instance a threshold of-120 dBm may be used. In this way, resources are not wasted making predictions based on low quality input data. If there are different output power for an inter-frequency or an inter-frequency cell as compared to the intra-frequency cell, the UE may be configured with a signal strength, signal quality offset. Similarly, the UE may be configured with the output power for each cell, inter-or intra-frequency. This may be very important if for instance different cell sites use different output powers. Without this it is likely that the prediction will be in error if the output power is different. The offset may also be configured for other different reasons, such as difference in antenna characteristics etc. Other prediction requirements related to the UE configuration may be: Authorization to enquire the inputs and post the outputs. For instance such requirement is important when attackers attempt to gain unauthorized access to the UE target environment where the model output can be modified or altered. Allocate the required UE central processing unit / graphical processing unit, "CPU / GPU", processing and storage / memory resources to perform the inference using the AI / ML model. For instance, allocations are to be made such that the AI / ML model will not overload the UE computing and storage resources. AI / ML model metrics such as accuracy, precision and recall, etc. For instance, if the Al model has a lower accuracy than a specific threshold, the UE will terminate the prediction or stop using the model, or stop from performing the predication or inference in future. In one aspect of the invention, the network configures a measurement object specifically for cells that are predicted instead of measured. In one aspect of the invention, as part of the measurement configuration a new type of object is introduced: a prediction object. The prediction object specifies how a prediction shall be performed; what inputs may be used to perform the prediction; model type; model size; model hyper-parameters such as e.g. learning rate; model metrics such as e.g. abstract performance range that defines the minimum and maximum performance as expressed on an abstract performance index; and policies such as e.g. activation / deactivation of the Al ML model. The input to a prediction may for instance be configured to be specific measurement object, or specific cells that have been configured in a measurement object. The prediction object can be tied to specific measurement objects, where the measurement object is derived from actual measurements performed and also be tied to measurement reporting configurations for reporting the measurement predictions. The prediction object may consist of one or more of the following configurable parameters for AI / ML: Model Performance: o Performance metric: the model performance metrics provide the quantitative metrics for evaluating a machine learning model, such as e.g. cost function value for both, regression and classification or Accuracy / Precision / Recall / F1-score for classification only. ML Model: o Inference type: Machine learning inferences can be broadly categorized into five classes: supervised learning, unsupervised learning, semi-supervised learning, self-supervised and reinforcement learning. o ML entity version. o Expected runtime / iterations. o Maximum runtime or iterations. ML model condition: o Updatable inference parameters. For instance, if one or more of the hyperparameters such as e.g. learning rate, has been updated for better performance, then the AI / ML inference model will have the new value to use. o Terminating inference conditions. For instance, if the accuracy threshold has been changed and the existing model cannot achieve the updated threshold, then the inference procedure will be stopped. o Upgrading model. For instance, if we change a Deep Neural Network, "DNN", model for a Deep Reinforcement Learning, "DRL", model. The prediction object may be associated with multiple measurement objects, for instance via a measurement object ID being configured. This can be used as input for prediction. This would for instance allow the UE to be configured to use measurements from multiple measurement objects in order to perform prediction. For instance, a prediction object may be configured with measurement objects on multiple frequencies, which are then combined for a single prediction in the prediction object, which may then be reported via a reporting configuration. This may also mean that a reporting configuration is associated with a prediction object in order for the prediction to be configured to be report. This can be done by introducing a prediction object ID to uniquely identify a prediction object. This can then be used to associate a reporting configuration with a prediction object. For the measurement object, or the prediction object, specific inter or intra-frequency cells may be configured where the UE may perform predictions. In an alternative aspect of the invention, the UE is configured with a measurement object with the option that the measurement object is only used for predictions. In another aspect of the invention, a new enhanced measurement identity may consist of an identity, a measurement object identity, a reporting configuration identity as well as a prediction identity. If a specific intra-frequency cell or frequency is configured with specific configuration, this can be inherited by the predicted cell, or considered to apply for the predicted cell. Specific configuration may for instance be: Measurement object offset. This is an offset that applies to all cells in a specific measurement object. Cell individual offset. An offset that is only for a specific cell in a measurement object. For instance, if the coverage quality of an inter-frequency cell is predicted based on measurements on an intra-frequency cell, whose intra-frequency cell has a cell individual offset, then the interfrequency cell will also apply the cell individual offset. Performing inter-frequency predictions In one aspect of the invention, the UE may perform the prediction / inference according to the following example: Step 2.1. The UE requests the model deployment from the UE storage, e.g. from Random Access Memory "RAM",) after it receives the gNB configuration request. For example, the inputs may be RSRP numerical values. Step 2.2. The deployed model executes inference / prediction using the available computing resources, for example using the UE CPU / GPU. In this way a three DNN layer model can be executed. The inference may be performed on a Layer 1 measurement or the Layer 3 filtered measurement of an intra or inter-frequency cell. It may also be performed based on a resulting final measurement report. Step 2.3. The output results from the execution are posted back to the UE storage. For example, the output is a categorical data type combined with probability value of occurrence. Step 2.4. The UE storage will compile and / or combine the results and AI / ML model configuration, before sending it back to the gNB. Reporting inter-frequency predictions In one aspect of the invention, the UE is configured to report the predicted coverage condition for an inter-frequency neighbour cell based on measurements of an intra-frequency cell, or another inter-frequency cell. The reported coverage can for instance be based on the signal strength, the signal quality or even the expected pathloss. This may be configurable. The reported quantity can be a coarse indication that indicates the quality of the coverage. This can be a small set of values. Such a quantity may be for instance be graded as a set of values including: ‘poor’, ‘good’; or ‘excellent’, ‘good’, ‘fair’, ‘poor’ The UE may also report the AI / ML model type that was used for the prediction. For instance, the UE may report that a DNN was used, or a linear or logistic regression was used. The UE may also report the accuracy, the prediction, the recall or the F1 score. In one aspect of the invention, the predicted coverage is included in measurement report that is triggered for measurements, for instance for intra-frequency measurements. This can be configured to be included in a measurement report. In other words, the predictions are piggybacked for measurement reports sent for triggered or periodical measurements. In one aspect of the invention, the UE can be configured to include the predicted coverage of the inter-frequency cell specific to the one or more cells that triggered the measurement report, or for instance predicted inter-frequency coverage of the cells in the list of the cells that triggered the measurement report. In another aspect of the invention, one or more inter-frequency cells that are of good quality, for instance based on a threshold, are reported in the measurement report. In one aspect of the invention, the UE indicates in the measurement report what input was used to predict the coverage of a cell. For instance, the UE may report that it used a specific intra-frequency cell, or an inter-frequency cell to perform the prediction. What is reported may for instance be one or more of: the cell ID, such as a Physical Cell ID, Cell ID or Global Cell ID of the cell used forthe prediction; the frequency of the cell used forthe prediction; the measurement object ID of the cell used for the prediction. The UE may also indicate multiple cells that were used forthe prediction. This has the benefit may not require that a UE is configured specifically to predict certain cells, but rather the UE may predict inter-frequency cells on its own, and the network may have to ignore certain predictions if there in fact are no inter-frequency cells on a specific frequency. This can for instance be useful in a handover situation where the UE may have measured an intra-frequency cell which is reported, and then the UE may further signal the predicted coverage of an inter-frequency cell associated with the intra-frequency cell. This can be used potentially forthe network to make a decision to handover to an inter-frequency cell. In one aspect of the invention, the prediction may trigger a report on its own. For instance, for a cell, which may be an inter-frequency cell, the prediction is triggered if it fulfils some type of condition, such condition may be a predetermined condition. The prediction may also be triggered periodically. Such measurement event may be termed a prediction measurement event. For instance, a prediction measurement event may be termed P1, or prediction measurement events termed P1, P2 and P3 etc. For instance, a prediction measurement event, such as a prediction measurement event here given the name P1, may be that a cell has a predicted measure over a certain configured threshold. For instance, another prediction measurement event, such as a prediction measurement event here given the name P2, may be that a cell has a predicted measure below a certain configured threshold. For instance, another prediction measurement event, such as a prediction measurement event here given the name P3 may be that a first cell has been predicted a measure larger than a that of a second cell. For instance, the second cell may be a cell whose measure has been predicted and not measured. The measured cell may for instance be an intra-frequency cell and the predicted cell may be an inter-frequency cell. The thresholds may be configured in the specific measure that the UE predicts, or that the UE is arranged to predict. To compare a predicted measure and a measured cell, there may be a need to ensure that they have the same quantity. In order to do this, the measured cell may be configured to be transformed to the same quantity as the predicted measures. For instance, if the predictions have the quantity ‘excellent’, ‘good’, ‘fair’ or ‘poor’, then the measured cell are transformed to one of these quantities. A specific reporting configuration may be introduced for reporting predictions. In one aspect of the invention, the UE is configured to report predictions. This can for instance be using the prediction events as explained above. This can for instance be a part of a specific prediction configuration, or it may be a part of a configuration which is a reporting configuration. The reporting configuration for reporting predictions may be associated with a prediction object, or it may be associated with a measurement object which is configured to be used for predictions. The UE may be configured in a reporting configuration to only report predictions above a certain threshold, where the threshold may be in the configured quantity. This can for instance be useful when the predictions are reported as part of a measurement report for measured cells. Although described with reference to inter-frequency predictions, which will be understood to have a specific technical meaning within the 5G NR framework, the invention as described may also be used for predicting inter-RAT measurements and as such the definitions and terminology relating to inter-frequency operations should be taken to include equivalent operations for handover for coverage / quality reasons or traffic steering. This is in particular important as inter-RAT measurements may consume even more power than those for inter-frequency operations. For inter-RAT measurements, the UE may be able to predict the coverage of another RAT by measuring the RAT the UE is currently operating or being connected to. For instance, the UE may be measuring characteristics of a 5G RAT and use this to predict an inter-RAT cell, where the inter-RAT cell may be a 4G Evolved Universal Terrestrial Radio Access Network "E-UTRAN" cell, or 6G RAT cell. An intra-RAT cell may be configured to be used for inter-RAT cell prediction and the specific RAT that may be predicted may also be configured by the network. The specific frequency of the RAT may also be configured. This can for instance be vital in a future 5G-6G deployment where there would be a need to measure a 5G and a 6G cell, but due to power consumption reasons it is preferred not to. In these cases it may be vital that a prediction regarding the 6G cell is made and reported to the network. INTRODUCTION TO THE DRAWINGS For a better understanding of the invention, and to show how embodiments of the same may be carried into effect, reference will now be made, by way of example, to the accompanying diagrammatic drawings in which: FIG. 1 shows steps in a background-art 5G NR handover procedure; FIG. 2 shows a high-level schematic representation of a background-art system for the enhanced mobility scenario; FIG. 3 shows a schematic representation of a typical inter-frequency handover scenario such as in the system of FIG. 2; FIG. 4 shows a signalling diagram for configuring and prediction in an inter-frequency produce according to an example embodiment; FIG. 5 shows intra-frequency cell configuration with the option to perform inter-frequency predictions of one or more inter-frequencies cells; FIG. 6 shows how a prediction object can be tied to specific measurement objects, where the measurement object is derived from actual measurements performed and also be tied to measurement reporting configurations for reporting the measurement predictions; FIG. 7 shows a signalling diagram for a method by which the UE may perform a prediction / inference according to an example embodiment; FIG. 8 shows an example for each of the UE prediction steps; FIG. 9 shows thresholds for a predicted cell and a measured cell, and triggering of a measurement report for a prediction event; FIG. 10 shows a schematic representation of a coverage based handover use case, according to an example embodiment; FIG. 11 shows a schematic representation of a traffic steering based handover use case, according to an example embodiment; and FIG. 12 shows a high-level schematic representation of a system for the enhanced mobility scenario incorporating a second embodiment for a traffic steering based handover use case. DESCRIPTION OF EXAMPLE EMBODIMENTS FIG. 4 shows a high-level signalling diagram for configuring and prediction in an inter-frequency produce according to an example embodiment. As illustrated, the three main components are firstly the gNB signalling to the UE to configure inter-frequency predictions. On receipt of the signal from the gNB, the UE performs inter-frequency predictions, Once the inter-frequency predictions have been performed the UE reports the inter-frequency predictions backtothegNB. FIG. 5 shows intra-frequency cell configuration with the option to perform inter-frequency predictions of one or more inter-frequencies cells. That is, to configure whether a UE shall perform predictions of an inter-frequency cell, there may be a number of options. One option is that an intra-frequency cell is configured with the option to perform inter-frequency predictions of one or more inter-frequencies. This can be useful in the case where not all cellsites have multiple frequencies cells. For instance, for some cells, the UE may not be configured to provide any inter-frequency predictions. In other options, as shown in FIG. 5, an intra-frequency cell is configured such that there may be inter-frequency predictions, and the specific frequencies where to perform predictions may be configured elsewhere, for instance it may be a general measurement configuration that applies to all cells. Examples of this for predictions in Frequency Ranges 1 and 2 are also shown in FIG. 5. FIG. 6 shows a prediction object. In particular, FIG. 6 shows how a prediction object can be tied to specific measurement objects, where the measurement object is derived from actual measurements performed and also be tied to measurement reporting configurations for reporting the measurement predictions. As set out above, as part of the measurement configuration a prediction object is introduced. The prediction object specifies how a prediction shall be performed, what inputs may be used to perform the prediction, model type, model size, model hyper-parameters, model metrics and policies. The input to a prediction may be a specific measurement object, such as shown in FIG. 6, or specific cells that have been configured in a measurement object. The prediction object can be tied to specific measurement objects such as shown for options A) and B) in FIG. 6. These are measurement objects derived from actual measurements performed and also be tied to measurement reporting configurations for reporting the measurement predictions. FIG. 7 shows a signalling diagram for a method by which the UE may perform a prediction / inference according to an example embodiment. That is, building on the signalling of FIG. 4, operations within the UE are performed to receive the signal for configuring interfrequency predictions from the gNB, perform computing operations to generate the predictions, and then transmit a report of the predictions back to the gNB. To deliver steps 2.2 through 2.5, a receive unit RX, a storage, a computing unit and a transmit unit TX are provided. FIG. 8 shows an example for each of the UE prediction steps, comprising first at 2.1 configuring an AI / ML model with inputs, e.g. parameters such as RSRP, RSRQ, and utilised PRB. These comprise a measurement report, such as for fi and PLM 'A', referencing Cells 1 through 5 and their respective RSRP. At 2.2 an inference model, for example a DNN, RL or similar is used to predict outputs in the form of inference results. The Al model shown comprises a three layer DNN. Then, at 2.3, the result of the prediction is stored, for subsequent reporting. The prediction results shown include those for operation at f2 for PLMN 'B', and comprise a ranking of signal strengths. FIG. 9 shows thresholds for a predicted cell and a measured cell, and triggering of a measurement report for a prediction event. The predictions may trigger a report on their own. For instance for a cell, which may be an inter-frequency cell, the prediction may be triggered if it fulfils some condition. FIG. 9 illustrates a triggering of a measurement event, termed a prediction measurement event.; The above general concepts are now put into context, by description of two specific handover examples. Example 1: Inter-frequency handover for coverage As shown in FIG. 10, the new procedure predicts the UEs expected coverage conditions, such as e.g. categorizes one or more of the RSRP, RSRQ, RSSI in four classes: excellent, good, fair and poor, for the inter-frequency neighbour, based on measurement report, and / or intra / inter-RAT information of the source gNB / cell frequency neighbours. When a UE sends an A2 report out, the UE can indicate if it falls into the coverage area of neighbour cells of another frequency. We propose a new information element in the A2 report for Prediction Output. This new information element may e.g. be named “predictedCoverage”. The Prediction Output is provided for each frequency, to indicate the coverage prediction for each target frequency within the A2 report. Upon receiving the A2 report from the UE, a gNB checks “predictedCoverage” for other frequencies and enables inter-frequency measurements only for those UEs where “predictedCoverage” for the target cell frequency is good. Example 2: Inter-frequency handover for traffic steering As shown in FIG. 11, in the case when traffic steering is required, i.e. Cell A radiating with fi is congested and some UEs will be offloaded to cells in f2, typically a significant portion of users in cell A are configured to measure f2 to understand if there is an available target cell in Freq B. One / some / all of these users are likely not in the f2 coverage area and so these users measure f2 unnecessarily, causing an additional overhead. We propose that gNB can ask UE to report on "predictedCoverage" for a list of frequencies and gets the update from UE before any inter-frequency measurement configuration update. Based on “predictedCoverage” information received from the UE, gNB decides which UEs to enable inter-frequency measurements. A proposed signalling diagram is depicted in FIG. 11. In the following discussion, parameters, interfaces, and procedures are described for one embodiment of the invention / implementation of the method, respectively. Taking the case of mobility enhancement in RRC_CONNECTED mode over air interface. A UE A is moving outside of Cell ‘A’ that operating using frequency fi and under PLMN A. Then the measurement values, e.g., RSRP, RSRQ, start to show need to trigger a Handover, "HO", procedure. However, the UE A performs a prediction measurements for a neighboured Cell ‘B’ that operating using frequency f2 and under PLMN B. The predicted measurements will be used in taking the decision for activating or de-activating the UE A inter-frequency measurements. For example, if the predicted f2 RSRP is 'Poor' for UE A, then the gNB A will not request the inter-frequency measurements from the UE A. However, when the predicted f2 RSRP is 'Excellent', then the gNB A will request UE A to start performing the inter-frequency measurements. FIG. 12 shows another embodiment of the invention, In FIG. 12, three UEs and two cells are shown. Cell A is shown as under PLMN A has high traffic load and serving two UEs: UE A and UE C using fi. Meanwhile, Cell B is shown as under PLMN B serving UE B using f2. UE A and UE C report the predicted measurements of f2 with ‘Poor’ and ‘Good’ RSRP to gNB A, respectively. In addition, gNB A status indicates a high traffic load caused by UE A. Therefore, gNB A sent a request only from UE C to start inter-frequency measurements before triggering the HO procedure for reducing the load of the gNB A. At the same time, the gNB B has low traffic load indicator and UE B with low Physical Resource Block, "PRB", demand and Excellent RSRP is reporting a predicted measurement of fi with 'Good' RSRP. As a result, the gNB B will not request activating the inter-frequency measurements from the UE B to prevent Cell A overloading. When the UE C HO procedure has been complete, gNB B will start receiving the predicted fi measurements while monitoring the Cell A traffic load percentage. Example use cases for the invention One of the examples for HO procedure requires activating both the A2 and A3 events for interfrequency measurements. Where Event A2 is triggered the UE may be configured with measurement gaps to measure the inter frequency objects and Event A3 for inter-frequency handover. In another scenario B1 and B2 events are used to move the UE from one cell to another for RAT load balance conditions instead of the RAT Radio conditions in a manner described in the document 3GPP TS 38.215 version 18.2.0 Release 18. For the Inter-frequency same RAT handover procedures use case: The known Inter-frequency measurements increase in specific RAT, causing a high UE energy consumption and reducing the user communication time. This is as described in 5G Measurement Events, https: / / www.techplayon.com / 5g-nr-measurement-events / (as accessed 25 September 2024. However, in example embodiments of this invention, the UE side AI / ML model reduces the number of exorbitant inter-frequency measurements that a UE needs to take by helping the serving gNB node to take the decision based on the predicted measurements for the neighbour cell frequency coverage. Only when the serving gNB takes the decision due to its own frequency measurements and / or the traffic load balancing conditions for one or more of the UEs within the serving cells, it will request a subset of the potential UEs that have a higher handover successful probability to start reporting the inter-frequency measurement using A3 event based on predicted coverage of the target frequency. As a result, the UEs with poor coverage prediction i.e. low probability of handover success procedure will not be measuring the inter-frequency cells and hence will not stop their communications and no extra UE energy consumption is required. For the Inter-frequency for inter-RAT handover procedures use case: In a similar use case to the one descripted before, example embodiments of this invention can be generalized for the inter-RAT scenario, where the events do not depends on the coverage. For example, Event B1 can be used for inter-RAT handover procedures, which only depends on load balancing feature, but not the coverage of the cell. In this case, if the invention reduces the Radio Link Failure, "RLF", as the prediction for the interfrequency will provide the indication of radio coverage before the Inter-RAT procedure of mobility starts. Which, similar to previous use case, will reduce the UE energy consumption and user communications down time. For example, Inter-RAT HO can be required if the UE moving out of 5G coverage and we may need IRAT measurements e.g. LTE, to handover. In capacity HO case, 5G cell can be congested and we may want to initiate a HO to LTE cell for capacity reasons. In both scenarios, the Al model will predict the LTE coverage and hence only those UEs with ‘Good’ LTE predicted coverage will be asked to measure the LTE frequency. That is, UEs with ‘Poor’ LTE predicted coverage will not be asked to measure inter-RAT frequency, saving on transmission gap2, UE battery, energy consumption etc. Specification examples, referred to as Example A and Example B are provided for the abovedescribed Measurement object configuration and for a Reporting Configuration, "ReportConfig" that dictates how and when a UE should send reports to the network. Examples A and B are drafted with reference to 3GPP TS 38.331 V18.3.0. EXAMPLE A — MeasObjectNR The IE MeasObjectNR specifies information applicable for SS / PBCH block(s) intra / inter-frequency measurements and / or CSI-RS intra / inter-frequency measurements. MeasObjectNR information element 1 loa / OEq ^ctHR / .-bFi-gnoiiO’: RRFOI-dluoHR 3 ubo h rrion 3p / o i ng; 33B-M7C ^SBRMTCl E0fFE0q'?3I-R3 ARFOH-VdUoHR Roforoio / oBignHl'lonfig 7hro / holdl.IR Ihro / holdllR nrof 33-71---0): / 700^0 / 0-30 ^2.. m.o::l)Eof 33-E 100): / 703..^000-30 ) nrof'?3l-R3~Rp / ouE00 / To3 / -pr0'30: ddidFdddSbdddiSddRiSRRddddddddTdAd ffootM0 Q-Off / otRongo7i / t; - / 011 / 7---70111---^071 st POI-Li / t OAOlOdAbdOdEOO Coll / ToAddl lodLi / d OdARanodERdO oi:oludodCollo7oAddl L'dLi / t 011-- / ^:10011 / 7--^0111--^071:^ allo-wodOoll / ToAddl L-dLi / t Ron T-El-m-nt ' , :0111 / 7:101 ( Fill (1. . iu / ::Rrof POI-Range / ) ) 01 E'CI- fEogEandlndioatotHR FroqB / indlndio / itOEHR nioo / OyoloROoll: M1OI, .f_rr, ofrl_, cf'd, cfl"_4 / nitoHi / t-Elo 33B-I I7d7i / t-rlc MeasGapId-rll assori at edl-IeasGapCSIRG-rll shiti41iit~rl HeasGapId-rlsi 00EH ITC4Li^t-rl“ [in 1' ( g hi. 0 r u r hi- in. rl^, in '4”, hi. 1’ ‘„4 idOOiGdAdOddO:S:£E^ CellcTv^ddHadLi-stEat-''’! I1 HeasGapId-rlu assneiat edl IeasGapCEIR31-vl"72i? ooSsiOiBi 1 leasSequence-rl 0 uellsToAddH’_'dListE::t-”1 E Ou cellsToAddModListExt-vl900 CellsToAddModListExt-vl900 ilil CellsTv.addI IndList d. .uh 1 Ta 111ar ) ) CellsTcLeddl Ind (1. . marllrofCelll leas ) ) GaklOgAOMsOGGWiW CellsToAddlk-dListErt-”1 E Ou (1.. ntazilIrofCelll leas ) ) GOOldAdOddEiO^ GidlldTdAddMddi E'hya Gellld oelllndividualGffset Q-OffsetRaugeListi iiOiOiWlOOiOOOGliW 0 {rhtp,Ihup,lintar} ntn-P-laEiaationHL-rV {Ehcp,Ihcp,linear} FtlltT’-AddMudEtt-’-lf FC' ntn-lIeight"?urCellInf'?-rl S lITE-nei nhF mi' =lllnf _-11 IntdKFKdjETdaidtl'jriC'jMl'J-rl? I nt =rFr=qE'r=dict innCnnf ig-rl S' 011111010111111000: {true}i prtdictiRrilritnvFvnpuentien-rO (1. . iTAZInterFre-p) ) ARFCII* OllOR ----------------------------------Example of 38.331 V18.3.0 EXAMPLE B ----------------------------------Example of 38.331 V18.3.0 - ReportConfig NR ReportConfigNR information element Repnrtdnnf igllR :: = EjEjEEggg'EllOllllllllllllllllllllll seeeeeeeeeeeeOIIOIIIIIseeeeee eondToiggerConfig-rl c eli^EventTriggered^rlo r::T”Periodic al-r 17 llllllllllllllll ifiiiiiiiiii^ FeOadicalRoort Config, EventTriggerConfig, CondTiOggerConfig-rl6 f CLI-E'oriC'die al Report Conf ig-nl C CLIOventTlOgerConfig^rlt, RxTxPeriodic.al-rl" , 1H >nR ..OlOti^ati n-il :j: if E'-entTriggerConfig : :- lililillli report Int erval report Amo u n t EepoEtQuar.it ityCell mazzReportCella r eport Quant it yRSIndevee report AddlleighI leas iillilllli me aaRi-SI-Repor tConfig-rlb igigigigiglSgSRSOSgSRglS ReanpgIaLgIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII:g:::::::::::::::::::::::::::::::::::::::::::::::::::: i n e 1 u d e R T - M e a a - e 16 i n o 1 u d eT ?L AH - II e a e - e 16 ino 1 ude5enaoe-I Ie a e -e 16 iiililiiiiii O'aEceL'.oat ionRequeat - e17 yl:agxa;g;A;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;gg:g:g:g:g:g:g:g:g:<g:g:g:g:g:g:g: EepoEtQuar.it it yRel ay-ri'7 dlt:pdgg:gg;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;gg:g:g:g:g:g:g:g:g:g:g:g liiiliiiiiiii numbeEtifTEiggeringAella-rl >- y$<:etg:g:rg;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;g;gg:g:g:g:g:g:g:g:g:<g:g: oelllndividualOffretLiat-e1~ Cen individuals faetLiat*E18 STI events SSThEeeheldSr 12 events-SD-Thre a hold-rl 2 |||||||||||||||||||||||||| RRAtPA p {rl, ri, e4, rL, e16, S, eU4, infinity},; LlMeRRRepd'rtQURhtuRyy'LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLg:: HeaaRepertQuantity vises, - < I leaaRRl-I-RepoEtConf ig-El 6 --tTyR-l-dc- [ET-lIam-Lict-ili } t 3L-1 Ie.a a Report Qu.ant it y-rll vR'i RIVAL S .ma v^llR^ Lt ) ■ , SQHRIIVL (flit (1. . marllrofCellHeaa ) ) SLSleaaTriggerQuantity^rlr. SSAQ AL-HeaaTEiggerQuantity-rli CALIA? AL, RAAlAdlAOiWlO^^flSR^ EEEEpEdAA^AlElElElElElElEE report Interval reportAuigunt EAifiAiAyi 11111111111111111111111111 reportQuantitype11 ma:<ReportCells report Qu ant it yR3-Inde::ea niarHrC'f RS-IndereeT’-'Rep’-'rt ttt / AtA 31:11:1:1:1:1:1:1:1:1:1:1:1:1:1:1:1:1:1:1:1:11111111 inoludeEe amMe aeurement 8 uaeAllowedCellLiet ilAAOMliilBAO 1 K a.R^p -11 yu ant it j . 1 , LiITAAA. (1- - MiNrof IndeneaToReport) ii'AlAC,: meaaR33I-ReportConfig-rln AC i-AAAAA' {true}; p A Al A 3 £ O OO AAAWSsOMSf}^ i ii o 1 ud e 3 en a o r -Me a s - r 1 n ul-DelayValueConfig-rl1 r e p o r t AM d H e i g 111 e a 8 - r 16 lillillll!^ iiiiiiiiiii ul-EnueeaDel ayA_-nf ig-rl 7 CetupReleaae {ET-14ameList-rl6} QAAAi'Al., •••• : EAAupRelAOAiE^OSWNAoBlWfcdA^^ :SAiiUplAOAO^l}}O^iWEAEljS?OBlA:O^ {AAtAAOlAAAAEiQEEiOlAdlAA^A^^^ 3-tii};_F-l-a> - [ HL-E i- cE-lay'1 nfi i-il" } ' , ijarceLucationRegueat-rl7 reportQuant it yRelay-rln iiiiiiiiiii 3L-Me asReport Quant ity-rl c- 111 PredictionReportConfig sssss^^d^S^ShEsE^sSiiSissssssssssssssss sssssBegoBSMeasuKedSKSSssssssissssssssssss sssssgKe^ctjRepdKtjiHgShBeshdSdSBSSjssssssssRsRRSRangesS'g jssissba^dS^h^ShSS^isssssssssss^ :<:<:<:<:<excei:lentGoodFair^^ :<:<:<:<:<:rSj^^$hXeShOi:d<:<:<:<:<:<:<:<:<:<^ PredictReportQuantity-rl9 ::= ----------------------------------Example of 38.331 V18.3.0--------------------------------- At least some of the example embodiments described herein may be constructed, partially or wholly, using dedicated special-purpose hardware. Terms such as ‘component’, ‘module’ or ‘unit’ used herein may include, but are not limited to, a hardware device, such as circuitry in the form of discrete or integrated components, a Field Programmable Gate Array (FPGA) or Application Specific Integrated Circuit (ASIC), which performs certain tasks or provides the associated functionality. In some embodiments, the described elements may be configured to reside on a tangible, persistent, addressable storage medium and may be configured to execute on one or more processors. These functional elements may in some embodiments include, by way of example, components, such as software components, object-oriented software components, class components and task components, processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. Although the example embodiments have been described with reference to the components, modules and units discussed herein, such functional elements may be combined into fewer elements or separated into additional elements. Various combinations of optional features have been described herein, and it will be appreciated that described features may be combined in any suitable combination. In particular, the features of any one example embodiment may be combined with features of any other embodiment, as appropriate, except where such combinations are mutually exclusive. Throughout this specification, the term “comprising” or “comprises” means including the component(s) specified but not to the exclusion of the presence of others. Attention is directed to all papers and documents which are filed concurrently with or previous to this specification in connection with this application and which are open to public inspection with this specification, and the contents of all such papers and documents are incorporated herein by reference. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and / or all of the steps of any method or process so disclosed, may be combined in any combination, except combinations where at least some of such features and / or steps are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. Thus, unless expressly stated otherwise, each feature disclosed is one example only of a generic series of equivalent or similar features. The invention is not restricted to the details of the foregoing embodiment(s). The invention extends to any novel one, or any novel combination, of the features disclosed in this specification (including any accompanying claims, abstract and drawings), or to any novel one, or any novel combination, of the steps of any method or process so disclosed.

Claims

1. A method of operating a User Equipment, UE, arranged for communication with a wireless communication system that comprises UEs and a network of cells, the method comprising reporting a predicted coverage condition for an inter-frequency neighbour cell based on measurements of an intra-frequency cell, or another inter-frequency cell.

2. The method of claim 1, wherein: in response to receipt of a measurement configuration from the network, predicting the coverage condition for an inter-frequency neighbour cell based on measurements of an intra-frequency cell, or based on measurements of another interfrequency cell; and reporting the prediction result to the network.

3. The method of claim 2, wherein the received measurement configuration configures the UE to measure intra-frequency cells, and in response the UE predicts the coverage of the interfrequency neighbour cell; or wherein received measurement configuration configures the UE to measure a first inter-frequency cell, and in response the UE is configured to predict the coverage of an inter-frequency neighbour cell that comprises a second inter-frequency cell.

4. The method of claim 2 or 3, wherein the method comprises predicting the coverage condition for an inter-frequency neighbour cell based on measurements performed of more than one intra and inter-frequency cells.

5. The method of any preceding claim, wherein the measurement configuration comprises conditions for the UE on whether it is able to perform predictions and / or report a prediction to the network, the conditions including one or more of:whether the inter-frequency or inter-frequency cell is within a certain frequency range away from an intra-frequency cell.whether the inter-frequency or inter-frequency cell is in a certain frequency range whether a specific cell or frequency has been configured to perform predictions whether an inter-frequency prediction is considered reliable enoughwhether the cell is configured as part of a measurement objectwhether the inter-frequency cell that is being predicted has not been measured, or measured recently.

6. The method of any preceding claim, wherein the UE is configured to determine the signal strength or signal quality of the intra-frequency cell, and if the signal strength or signal quality of the intra-frequency cell is below a threshold, the UE does not attempt to perform a prediction or does not report a prediction.

7. The method of any preceding claim, comprising, configuring, by the network, a measurement object for cells that are predicted instead of measured, such measurement object referred to as a prediction object and specifying one or more of: how a prediction shall be performed; what inputs may be used to perform the prediction; prediction model type; prediction model size; prediction model hyper-parameters such as learning rate; prediction model metrics such as an abstract performance range that defines the minimum and maximum performance as expressed on an abstract performance index; and policies such as activation / deactivation of the AI / ML model.

8. The method of claim 7, wherein the input to a prediction is configured to be specific measurement object, or specific cells that have been configured in a measurement object, and the prediction object is tied to specific measurement objects, where the measurement objects are derived from actual measurements performed, as well as being tied to measurement reporting configurations for reporting the measurement predictions.

9. The method of claim 7 or 8, wherein the prediction object consists of one or more of the following configurable parameters for AI / ML:Model Performance;ML Model;ML model condition;10. The method of any preceding claim, wherein the UE performs the prediction according to the following steps:the UE requests AI / ML model deployment from its storage after it receives the measurement configuration;the deployed AI / ML model executes a prediction using the available computing resources of the UE;the output results from the execution are posted back to UE storage;the UE storage compiles and / or combine the results and AI / ML model configuration, before reporting back to the network.

11. The method of any preceding claim, wherein the reported predicted coverage condition is based on the signal strength, the signal quality or the expected pathloss, and indicates the quality of the coverage as one of a set of graded values.

12. The method of any preceding claim, wherein the predicted coverage is included in measurement report that is triggered for measurements, such that the predictions are piggybacked for measurement reports for triggered or periodical measurements.

13. The method of any preceding claim, comprising the UE reporting the predicted coverage of the inter-frequency cell, to one or more cells that triggered the measurement with a measurement condition.14, The method of any preceding claim, comprising the UE generating the predicted coverage condition using an AI / ML model, and comprising reporting to the network the AI / ML model type that was used for the prediction.

15. The method of claim 14, comprising the UE reporting one or more of: the accuracy of the prediction; the recall; or the F1 score.

16. The method of any preceding claim, comprising the UE only reporting predictions above a certain threshold.

17. The method of any preceding claim, comprising the UE reporting what input was used to predict the coverage of a cell.

18. The method of any preceding claim, wherein the inter-frequency neighbour cell comprises a cell on another RAT network.

19. The method of any preceding claim, wherein a UE operating using frequency fi on Cell 'A' under PLMN ‘A’ performs prediction measurements for a neighbored Cell ‘B’ that operates using frequency f2 and under a different PLMN, PLMN ‘B’.

20. Apparatus arranged to perform the method of any preceding claim.A

Images