Measurement reporting

Abstract

The invention comprises obtaining a reporting framework at a network device such as a UE or base station, where the reporting framework maps one or more accuracy conditions to one or more reporting behaviours such as reporting periodicity. Once an indication of whether a measurement meets an accuracy condition is obtained a change of reporting behaviour may be indicated and switched based upon the reporting framework and the indication of whether the accuracy condition is being met. The accuracy conditions may comprise an accuracy threshold and the accuracy may be expressed as a mean square error, negative logarithm or likelihood or expressed as a comparison to a ground truth determined using more accurate data.

Description

Field This specification relates to apparatuses and methods for determining reporting behaviour and apparatuses and methods for verifying accuracy conditions. Background There remains an interest in improved measurement reporting in Radio Access Networks. Summary The scope of protection sought for various embodiments of the invention is set out by the independent claims. The embodiments and features, if any, described in this specification that do not fall under the scope of the independent claims are to be interpreted as examples useful for understanding various embodiments of the invention. A first aspect provides an apparatus comprising: means for obtaining a reporting framework, wherein the reporting framework maps one or more accuracy conditions to one or more reporting behaviours and / or one or more changes of reporting behaviour; means for obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; and means for determining a reporting behaviour and / or one or more changes of reporting behaviour based at least in part upon the reporting framework and the indication of whether an accuracy condition of the one or more accuracy conditions is met by the output. In some example embodiments, the one or more accuracy conditions comprise at least one or more of the following: a threshold accuracy above which an accuracy condition is met; a threshold accuracy below which an accuracy condition is met; an accuracy range within which an accuracy condition is met; and an accuracy range outside of which an accuracy condition is met. In some example embodiments one or more accuracies of the one or more accuracy conditions are defined in terms of at least one or more of: a proportion of one or more outputs that match a one or more corresponding ground truths; a mean square error of one or more outputs with respect to one or more corresponding ground truths; and one or more negative logarithms of one or more likelihoods, wherein the likelihoods are likelihoods assigned to outputs that match corresponding ground truths. In some example embodiments the output comprises one or more of: a predicted value of a measurement; a target cell; and a target beam. In some example embodiments the one or more reporting behaviours each comprise the use of a respective measurement report configuration. The respective measurement report configurations of the one or more reporting behaviours may each comprise at least one or more of at least the following parameters: a reporting criteria; an event that is to trigger reporting; a report period / interval; and an amount of reports. The respective measurement report configurations of the one or more reporting behaviours may comprise at least one or more accuracy levels at which one or more user device measurements are to be collected. A first reporting behaviour of the one or more reporting behaviours may comprise a first measurement report configuration comprising a first accuracy level at which one or more user device measurements are to be collected; a second reporting behaviour of the one or more reporting behaviours may comprise a second measurement report configuration comprising a second accuracy level at which one or more user device measurements are to be collected, wherein the first accuracy level is higher than the second accuracy level; a first accuracy condition may be met above a threshold and map to the second reporting behaviour; and a second accuracy condition may be met below a threshold and map to the first reporting behaviour. In some example embodiments the reporting behaviours comprise event triggered measurement reports and the reporting framework comprises one or more measurement report configurations mapping the one or more accuracy conditions to triggers of event triggered measurement reports. In some example embodiments the apparatus further comprises means for collecting measurement data associated with radio access network signals; the means for obtaining the reporting framework comprises means for receiving the reporting framework from a network node; and the means for obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions comprises means for generating the output based on measurement data obtained by the user device, means for determining an accuracy of the generated output, and means for verifying whether the generated output meets the accuracy condition. The one or more reporting behaviours may each comprise the use of a respective measurement report configuration, and the apparatus may further comprise: means for, responsive to obtaining an indication that the output meets an accuracy condition, sending a request to a network node to use a measurement report configuration comprised by the reporting behaviour mapped to that accuracy condition by the reporting framework; means for receiving from the network node a response indicating that the measurement report configuration is to be used by the apparatus; and means for applying the indicated measurement report configuration. The accuracy condition met by the output may be mapped to a reporting behaviour comprising a measurement report configuration having an event-based reporting criteria; and the apparatus may further comprise: means for verifying that an event identified in the measurement configuration has occurred; and means for triggering a measurement report based on verifying that the event has occurred. The apparatus may further comprise means for sending to the network node capability information associated with the user device; and means for sending to the network node accuracy requirements for the output. In some example embodiments the apparatus is a network node, and: the apparatus further comprises means for receiving measurement data associated with radio access network signals from a user device; the means for obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions comprises means for generating the output based on the measurement data received from the user device, means for determining an accuracy of the generated output, and means for verifying whether the generated output meets the accuracy condition. The apparatus may further comprise means for sending to the user device an instruction to use a measurement report configuration comprised by the reporting behaviour mapped to the met accuracy condition. The apparatus may further comprise means for receiving from the user device capability information; and the means for obtaining a reporting framework may further comprise means for generating a reporting framework based on capability information associated with a user device and accuracy requirements for the output. In some example embodiments, the apparatus is a network node, and wherein: the means for obtaining a reporting framework comprises means for generating a reporting framework based on capability information associated with a user device and accuracy requirements of outputs of a user device; the apparatus comprises means for sending the one or more accuracy conditions to the user device; the apparatus comprises means for receiving from a user device an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; and the apparatus comprises means for sending to the user device an instruction to use a measurement report configuration comprised by the reporting behaviour mapped to the met accuracy condition. A second aspect provides a method comprising: obtaining a reporting framework, wherein the reporting framework maps one or more accuracy conditions to one or more reporting behaviours and / or one or more changes of reporting behaviour; obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; and determining a reporting behaviour and / or one or more changes of reporting behaviour based at least in part upon the reporting framework and the indication of whether an accuracy condition of the one or more accuracy conditions is met by the output. In some example embodiments, the one or more accuracy conditions comprise at least one or more of the following: a threshold accuracy above which an accuracy condition is met; a threshold accuracy below which an accuracy condition is met; an accuracy range within which an accuracy condition is met; and an accuracy range outside of which an accuracy condition is met. In some example embodiments one or more accuracies of the one or more accuracy conditions are defined in terms of at least one or more of: a proportion of one or more outputs that match a one or more corresponding ground truths; a mean square error of one or more outputs with respect to one or more corresponding ground truths; and one or more negative logarithms of one or more likelihoods, wherein the likelihoods are likelihoods assigned to outputs that match corresponding ground truths. In some example embodiments the output comprises one or more of: a predicted value of a measurement; a target cell; and a target beam. In some example embodiments the one or more reporting behaviours each comprise the use of a respective measurement report configuration. The respective measurement report configurations of the one or more reporting behaviours may each comprise at least one or more of at least the following parameters: a reporting criteria; an event that is to trigger reporting; a report period / lnterval; and an amount of reports. The respective measurement report configurations of the one or more reporting behaviours may comprise at least one or more accuracy levels at which one or more user device measurements are to be collected. A first reporting behaviour of the one or more reporting behaviours may comprise a first measurement report configuration comprising a first accuracy level at which one or more user device measurements are to be collected; a second reporting behaviour of the one or more reporting behaviours may comprise a second measurement report configuration comprising a second accuracy level at which one or more user device measurements are to be collected, wherein the first accuracy level is higher than the second accuracy level; a first accuracy condition may be met above a threshold and map to the second reporting behaviour; and a second accuracy condition may be met below a threshold and map to the first reporting behaviour. In some example embodiments the reporting behaviours comprise event triggered measurement reports and the reporting framework comprises one or more measurement report configurations mapping the one or more accuracy conditions to triggers of event triggered measurement reports. In some example embodiments: the method further comprises collecting measurement data associated with radio access network signals; obtaining the reporting framework comprises receiving the reporting framework from a network node; and obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions comprises generating the output based on the collected measurement data, determining an accuracy of the generated output, and verifying whether the generated output meets the accuracy condition. The one or more reporting behaviours may each comprise the use of a respective measurement report configuration, and the method may further comprise: responsive to obtaining an indication that the output meets an accuracy condition, sending a request to a network node to use a measurement report configuration comprised by the reporting behaviour mapped to that accuracy condition by the reporting framework; receiving from the network node a response indicating that the measurement report configuration is to be used by the apparatus; and applying the indicated measurement report configuration. The accuracy condition met by the output may be mapped to a reporting behaviour comprising a measurement report configuration having an event-based reporting criteria; and the method may further comprise: verifying that an event identified in the measurement configuration has occurred; and triggering a measurement report based on verifying that the event has occurred. The method may further comprise sending to the network node capability information; and sending to the network node accuracy requirements for the output. In some example embodiments: the method further comprises receiving measurement data associated with radio access network signals from a user device; obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions comprises generating the output based on the measurement data received from the user device, determining an accuracy of the generated output, and verifying whether the generated output meets the accuracy condition. The method may further comprise sending to the user device an instruction to use a measurement report configuration comprised by the reporting behaviour mapped to the met accuracy condition. The method may further comprise receiving from the user device capability information; and obtaining a reporting framework may comprise generating a reporting framework based on capability information associated with a user device and accuracy requirements for the output. In some example embodiments, obtaining a reporting framework comprises generating a reporting framework based on capability information associated with a user device and accuracy requirements of outputs of a user device; the method comprises sending the one or more accuracy conditions to the user device; the method comprises receiving from a user device an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; and the method comprises sending to the user device an instruction to use a measurement report configuration comprised by the reporting behaviour mapped to the met accuracy condition. A third aspect of provides a computer program comprising a set of instructions which, when executed on an apparatus, is configured to cause the apparatus to carry out a method comprising: obtaining a reporting framework, wherein the reporting framework maps one or more accuracy conditions to one or more reporting behaviours and / or one or more changes of reporting behaviour; obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; and determining a reporting behaviour and / or one or more changes of reporting behaviour based at least in part upon the reporting framework and the indication of whether an accuracy condition of the one or more accuracy conditions is met by the output. In some example embodiments, the third aspect may include any other feature mentioned with respect to the method of the second aspect. A fourth aspect of the invention provides a non-transitory computer-readable medium having stored thereon computer-readable code, which, when executed by at least one processor, causes the at least one processor to perform a method, comprising: obtaining a reporting framework, wherein the reporting framework maps one or more accuracy conditions to one or more reporting behaviours and / or one or more changes of reporting behaviour; obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; and determining a reporting behaviour and / or one or more changes of reporting behaviour based at least in part upon the reporting framework and the indication of whether an accuracy condition of the one or more accuracy conditions is met by the output. The fourth aspect may include any other feature mentioned with respect to the method of the second aspect. A fifth aspect of the invention provides an apparatus, the apparatus having at least one processor and at least one memory having computer-readable code stored thereon which when executed controls the at least one processor to: obtain a reporting framework, wherein the reporting framework maps one or more accuracy conditions to one or more reporting behaviours and / or one or more changes of reporting behaviour; obtain an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; and determine a reporting behaviour and / or one or more changes of reporting behaviour based at least in part upon the reporting framework and the indication of whether an accuracy condition of the one or more accuracy conditions is met by the output. The fifth aspect may include any other feature mentioned with respect to the method of the second aspect. A sixth aspect provides an apparatus comprising: means for obtaining user device measurement data; means for processing user device measurement data to determine an output; means for determining an indication of an accuracy of the output; and means for comparing the indicated accuracy with one or more accuracy conditions. In some example embodiments, the determined indication of an accuracy of the output and the one or more accuracy conditions are defined in terms of at least one or more of: a proportion of determined outputs that match a ground truth; a mean square error of determined outputs compared to a ground truth; and a negative log likelihood. In some example embodiments, the apparatus further comprises means for determining a ground truth, wherein the means for determining an indication of an accuracy of the output is configured to determine the indication of accuracy based at least in part on the output and a determined ground truth, and wherein the means for determining a ground truth is configured to determine a ground truth based at least in part on one or more of: second measurement data, wherein the second measurement data is more accurate than the user device measurement data; a processing method that Is more computationally intensive than a processing method used by the means for processing user device measurement data to determine the output; third measurement data, wherein third measurement data comprises more data than user device measurement data; and fourth measurement data, wherein fourth measurement data comprises data collected after the user device measurement data was collected. In some example embodiments, the one or more accuracy conditions comprise at least one or more of: a threshold accuracy above which an accuracy condition is met, a threshold accuracy below which an accuracy condition is met, an accuracy range within which an accuracy condition is met, and an accuracy range without which an accuracy condition is met. In some example embodiments, the user device measurement data comprises measurement data indicating one or more of: a reference signal received power; a reference signal received quality; a received signal strength indicatior; a signal to interference plus noise ratio; and a signal to noise ratio. In some example embodiments, the output comprises one or more of: a target cell; and a target beam. In some example embodiments, the output comprises one or more predicted values of a measurement. In some example embodiments, the apparatus is a user device, and the apparatus further comprises means for collecting the user device measurement data. The means for collecting the user device measurement data may be configured to collect data at different levels of accuracy. In some example embodiments the apparatus is a network node, and the apparatus comprises means for receiving the user device measurement data from a user device. A seventh aspect provides a method comprising: obtaining user device measurement data; processing user device measurement data to determine an output; determining an indication of an accuracy of the output; and comparing the indicated accuracy with one or more accuracy conditions. In some example embodiments, the determined indication of an accuracy of the output and the one or more accuracy conditions are defined in terms of at least one or more of: a proportion of determined outputs that match a ground truth; a mean square error of determined outputs compared to a ground truth; and a negative log likelihood. In some example embodiments the method further comprises determining a ground truth, wherein determining an indication of an accuracy of the output comprises determining the indication of accuracy based at least in part on the output and a determined ground truth, and wherein determining a ground truth comprises determining a ground truth based at least in part on one or more of: second measurement data, wherein the second measurement data Is more accurate than the user device measurement data; a processing method that is more computationally intensive than a processing method used in the step of processing user device measurement data to determine the output; third measurement data, wherein third measurement data comprises more data than user device measurement data; and fourth measurement data, wherein fourth measurement data comprises data collected after the user device measurement data was collected. In some example embodiments the one or more accuracy conditions comprise at least one or more of: a threshold accuracy above which an accuracy condition is met, a threshold accuracy below which an accuracy condition is met, an accuracy range within which an accuracy condition is met, and an accuracy range without which an accuracy condition is met. In some example embodiments user device measurement data comprises measurement data indicating one or more of: a reference signal received power; a reference signal received quality; a received signal strength indicator; a signal to interference plus noise ratio; and a signal to noise ratio. In some example embodiments the output comprises one or more of: a target cell; and a target beam. In some example embodiments the output comprises one or more predicted values of a measurement. In some example embodiments the method further comprises collecting the user device measurement data. Collecting user device measurement data may comprise collecting data at different levels of accuracy. In some example embodiments the method further comprises receiving the user device measurement data from a user device. An eighth aspect of provides a computer program comprising a set of instructions which, when executed on an apparatus, is configured to cause the apparatus to carry out a method comprising: obtaining user device measurement data; processing user device measurement data to determine an output; determining an Indication of an accuracy of the output; and comparing the indicated accuracy with one or more accuracy conditions. In some example embodiments, the eighth aspect may include any other feature mentioned with respect to the method of the seventh aspect. A ninth aspect of the invention provides a non-transitory computer-readable medium having stored thereon computer-readable code, which, when executed by at least one processor, causes the at least one processor to perform a method, comprising: obtaining user device measurement data; processing user device measurement data to determine an output; determining an indication of an accuracy of the output; and comparing the indicated accuracy with one or more accuracy conditions. The ninth aspect may include any other feature mentioned with respect to the method of the seventh aspect. A tenth aspect of the invention provides an apparatus, the apparatus having at least one processor and at least one memory having computer-readable code stored thereon which when executed controls the at least one processor to: obtain user device measurement data; process user device measurement data to determine an output; determine an indication of an accuracy of the output; and compare the indicated accuracy with one or more accuracy conditions. The tenth aspect may include any other feature mentioned with respect to the method of the seventh aspect. Brief Description of the Drawings Example embodiments will now be described by way of non-limiting example, with reference to the accompanying drawings, in which: FIG. 1 is diagram illustrating schematically an RRC message structure of a measurement configuration in accordance with example embodiments; FIG. 2 is a diagram showing a message flow sequence in accordance with example embodiments; FIG. 3 is a graph illustrating at a high level how accuracy-based event entering and exit conditions may be used in event triggered reporting in accordance with example embodiments; FIGS. 4-8 are diagrams showing message flow sequences in accordance with example embodiments; FIG. 9 is a block diagram showing an apparatus in accordance with example embodiments. FIGS. 10 and 11 are flow diagrams showing methods in accordance with example embodiments; FIG. 12 is a diagram showing schematically an implementation of a ML time-series prediction model in accordance with example embodiments. FIG. 13 is a block diagram of components of a system in accordance with an example embodiment; and FIG. 14 shows an example of tangible media for storing computer-readable code which when run by a computer may perform methods according to example embodiments described above. Detailed Description In a cellular Radio Access Network (RAN), a user equipment (UE) may be served by a cell. It may become desirable for the user device to be served by a different cell, and so the network may execute a handover procedure, to change the cell serving the UE. A determination that a handover to another cell is to take place may be made at a network node of the RAN based on measurements. All or a portion of these measurements may be made at the user device, and these measurements may be communicated to a network node in one or more measurement reports. There may therefore be an interest in suitable measurements being made at a UE, and in suitable information being sent to a network node, so that appropriate handover decisions may be made. Other decisions by the user device and / or network may also rely on suitable UE measurements and measurement reports. The UE and network may take a "static" approach to making and reporting measurements and making handover decisions on the basis of those reports. Under this static approach, high-fidelity measurements may be required to ensure that accurate handover decisions are made. Measurement configuration The measurements to be made by a user device and how these are communicated to a RAN node may be specified in a measurement configuration stored by the UE. A UE may be provided with a measurement configuration via a Radio Resource Control (RRC) message, provided by a network node. Figure 1 is diagram illustrating schematically an RRC message structure of a measurement configuration 100. Measurement configuration 100 comprises an indication of one or more measurement objects 110, one or more reporting configurations 112, an indication of one or more measurement identities 114, one or more quantity configurations 116, and an indication of one or more measurement gaps 118. Measurement object 110 is an entity on which the UE performs measurements. Measurement objects may vary by the type of measurement, such as intra-frequency, inter-frequency, and inter-RAT (Radio Access Technology) measurements. Measurement objects may include details such as target cell frequencies, locations in frequency or time, target reference signals, cell-specific offsets, and lists of blacklisted / whitelisted cells. A measurement object may comprise an object ID that links the measurement object to one or more serving cells. Reporting configuration 112 defines how the UE reports measurements. Reporting configurations may include the criteria for triggering a report (e.g., an event that may trigger a measurement report in the case of event-based reporting). Reporting configurations may include an indication of the reference signal (RS) type used for measurements. For example, the reference signal may be a sync signal (SSB) / physical broadcast channel (PBCH) signal, or the reference signal may be a Channel State Information (CSI) Reference Signal (CSI-RS). Reporting configurations may include a reporting format. Conditional reconfiguration includes additional criteria and reference signal types for execution. The one or more measurement identities 114 are lists linking measurement objects to reporting configurations. Multiple measurement identities can link to one reporting configuration, and the measurement Identity is included in the report sent to the network. The one or more quantity configurations 116 are configurations determining the filtering applied to measurements during event evaluation, reporting, and periodic reporting. They can have different filter coefficients for various measurement quantities and RS types. The one or more measurement gaps 118 are time periods that the UE can utilize to perform the specified measurements. For each measurement object, one or more reporting configurations can be defined. There can be one or more reporting configurations per measurement object. Each reporting configuration may comprise one or more of a reporting criterion, an RS type, and a reporting format. When the reporting criterion is determined to be met, this may trigger the UE to send a measurement report. The reporting criterion may either be periodic (i.e., configured so that the UE sends a measurement report periodically) or event triggered (e.g., single event triggered). The RS type refers to the reference signal that the UE uses for beam and cell measurement results (SSB or CSI-RS). The reporting format comprises the quantities per cell and per beam that the UE includes in the measurement report (e.g. RSRP), and may comprise other associated information such as the maximum number of cells and the maximum number beams per cell to report. LI CSI reporting CSI reporting may be carried out at the physical layer (LI). In physical layer CSI reporting, the UE may calculate CSI parameters (if reported) assuming the following dependencies between CSI parameters (if reported): LI may be calculated conditioned on the reported CQI, PMI, RI and CRI; CQI may be calculated conditioned on the reported PMI, RI and CRI; PMI may be calculated conditioned on the reported RI and CRI; and RI may be calculated conditioned on the reported CRI. The Reporting configuration for CSI can be aperiodic, and the report may be sent using the Physical Uplink Shared Channel (PUSCH). The Reporting configuration can be periodic, and the report may be sent using the Physical Uplink Control Channel (PUCCH). The Reporting configuration can be semi-persistent, and the report may be sent using the PUCCH, and the downlink control information (DCI) activated PUSCH. The CSI-RS Resources can be periodic, semi-persistent, or aperiodic. Table 1 below shows combinations of CSI Reporting configurations and CSI-RS Resource configurations, and how the CSI Reporting may be triggered (or activated) for each CSI-RS Resource configuration. Periodic CSI-RS may be configured by higher layers. Table 1 - Triggering of potential CSI Reporting types for CSI-RS Configurations CSI-RS Configuration Periodic CSI Reporting Semi-Persistent CSI Reporting Aperiodic CSI Reporting Periodic CSI- RS No dynamic triggering For reporting on PUCCH, the UE receives an activation command; for reporting on PUSCH, the UE is triggered by DCI Triggered by DCI. Semi- Persistent CSI- RS Not Supported For reporting on PUCCH, the UE receives an activation command; for reporting on PUSCH, the UE is triggered by DCI Triggered by DCI. Aperiodic CSI-RS Not Supported Not Supported Triggered by DCI. Life Cycle Management UE measurements may be used in Radio Resource Management (RRM). UE RRM measurements may be input data for algorithms, which may implement machine learning (ML) models. These algorithms may implement models designed to predict RRM measurement and mobility-related events. ML models are trained models, and a "lifecycle" of such models may be managed In Life Cycle Management (LCM) procedures. LCM may include the following aspects: Data collection (this may also Include associated assistance information, if applicable); Model training; Functionality / model identification; Model transfer; Model inference operation; Functionality / model selection, activation, deactivation, switching, and fallback operation (this may include decisions by the network (either network initiated or UE-initiated and requested to the network), and / or decisions by the UE (event-triggered as configured by the network, UE's decision reported to the network, or UE-autonomous either with UE's decision reported to the network or without it); Functionality / model monitoring; Model update; and UE capability An ML model may have a model ID with associated information. An ML model may also provide given functionality. Different models may have different requirements such as measurement accuracy and prediction accuracy. A flexible (measurement) reporting mechanism may suit different ML requirements while performing LCM procedures at the UE side and / or the network side. When deploying ML solutions for mobility (handover decision making) one or both of the network (NW) and the user equipment (UE) may be equipped with Al-driven (or ML-driven) algorithms. The accuracy of the predictions made by such algorithms may hinge on the nature of the model inputs. For example, algorithms for inter-cell Beam-level measurement prediction for network layer (L3) mobility may require high-fidelity measurements as inputs, to ensure the high reliability of prediction outcomes, i.e., target cell and / or target beam. Varying the measurement accuracy for the device can be challenging for several reasons: • Implementing high-fidelity measurements may introduce complexity; • Performing high-accuracy measurements may require high processing overhead, higher power consumption, activation of additional RF components, etc.; • Reporting measurements with different accuracies might not match existing requirements, e.g., accuracy levels of RSRP reporting; and • For AI / ML based use cases, e.g., beam management or mobility, UE can support different prediction accuracy levels depending on the conditions and scenarios. The static measurement reporting procedure may not capture requirement of the AI / ML based prediction accuracy levels in a given use case. Furthermore, there are different radio conditions and scenarios when higher accuracy of measurements may be needed or not, for example: • When UE Is In good coverage of sufficient number of beams / cells, such as Line of Sight (LoS) conditions, high accuracy measurement may not be needed, and predictions may be reliable enough. However, in strong Non Line of Sight (NLoS) conditions, or when many beams are blocked, increased measurement accuracy may help provide reliable predictions despite inference. • When UE is in high mobility state, cell-edge, non-statlonary status, etc, higher accuracy measurements may be needed. • Different LCM aspects may require different measurement accuracy levels. In one example, the data collection for model training might require high fidelity measurements. High fidelity measurements may also be required in ground truth data collection for model monitoring (e.g., monitoring of the accuracy of predictions made by the model). • While monitoring, a degradation of the inference may be observed based on the default / low accuracy measurements. For example, it may be observed that low accuracy measurements result in sufficiently accurate model predictions / outputs. • Performing high-fidelity measurements and / or reporting high-fidelity measurement data can provide data that can be used as ground truth for monitoring purposes at the UE and / or NW side. • Different AI / ML models / model inputs might be needed or used when different measurement accuracies are assumed. Thus, mechanisms to identify a capability for measuring and / or reporting measurements with different accuracy or precision levels, and report using, switch between, and test using different levels of measurement accuracy and or precision may be useful. In AI / ML use cases, these mechanisms may be useful. A dynamic measurement and reporting mechanism based on various (e.g., two or more) measurement accuracy levels, which may be useful for AI / ML use cases, may comprise one or more of the following aspects. UE-sided embodiment In a UE-sided embodiment, a UE hosts an algorithm that makes use of measurement data collected by the UE to produce an output. In some example embodiments, this algorithm may be an AI / ML based algorithm. The UE monitors accuracy requirements of the output of the algorithm. The output of the algorithm may be part of the content of, or used to determine the content of, a measurement report sent to the network by the UE. In this embodiment, UE may send capability information to a network node (such as a gNB) to indicate its supported reporting mechanism(s) and parameters, and the network node may provide a corresponding reporting configuration for the UE to execute a dynamic reporting framework based on an accuracy requirement (which may be an accuracy requirement of the output). The reporting configuration may comprise a reporting configuration using a periodic, event triggered, or semi-persistent / aperiodic reporting mechanism. The selected reporting configuration may be based on the indicated supported reporting mechanism. NW-sided embodiment In a NW-sided embodiment, the NW (and in some examples a particular network node) hosts an algorithm that makes use of measurement data collected by the UE to produce an output. In some example embodiments, this algorithm may be an AI / ML based algorithm. The NW monitors accuracy requirements of the output of the algorithm, and may indicate to the UE to report measurements having an accuracy or accuracy level. In this embodiment, based on one or more accuracy requirements at the NW side, the NW may provide a desired configuration for the UE to execute this dynamic reporting framework. Similarly, the reporting configuration may comprise a reporting configuration using a periodic, event triggered, or semi-persistent / aperiodic reporting mechanism. The selected reporting configuration may be based on the indicated supported reporting mechanism. One or both of the above referenced embodiments may use one or more configuration messages including the one or more of the following aspects to provide dynamic accuracybased reporting: • UE capability information, which may include the reporting mechanisms supported by the UE, and which may be sent from the UE to the network. • RRC configuration parameters for the dynamic accuracy-based reporting mechanism, which may include report criteria, report amount, interval, etc., which may be sent from the network to the UE. • Accuracy verification conditions and thresholds for the UE and / or the NW to evaluate accuracy requirements. In periodic reporting and semi-persistent reporting, the accuracy verification conditions and / or thresholds may be conditions for changing between measurement and / or reporting configurations having different measurement and / or reporting accuracy levels. • One or more time windows and or durations for the UE and / or the NW to evaluate constraints (e.g., accuracy verification conditions and thresholds). • One or more accuracy based events configurations for the UE and / or the NW to perform the event triggered reporting. The configuration may include event start and exit conditions and associated thresholds. Figure 2 is a diagram showing message flow sequence 200. Message flow sequence 200 takes place between UE 202 and network node 204. Network node 204 may in some embodiments be a gNB. Message flow sequence 200 corresponds to an embodiment in which an algorithm that determines an output is executed at the UE side. The algorithm may be an ML algorithm. The accuracy of the output is determined and verified against an accuracy condition. On the basis of this verification, the reporting configuration used by UE 202 may be changed. Message flow sequence 200 depicts changing the reporting configuration to a reporting configuration having a periodic reporting type. At step 210, network node 204 sends a capability enquiry to UE 202 for performing the dynamic measurement reporting procedure. The enquiry may be sent via RRC signalling. At step 212, UE 202 sends capability information to network node 204. This capability information may be sent alongside measurement accuracy requirements for an output determined at UE 202. This message may be sent via RRC signalling. The supported reporting frameworks (and / or the supported reporting configurations) and the corresponding parameters, such as supported report method types, report periodicities, amount of reports, etc., may be included in this message as capability information. At step 214, network node 204 prepares configuration parameters according to the capability information to create a reporting framework. This reporting framework may be prepared for use in cases where UE 202 uses AI / ML to determine an output, such as RRM measurement prediction (e.g., for L3 mobility). At step 216, network node 204 sends a (re)configuration message to UE 202 with the configuration parameters to support the new reporting framework e.g., type of report, periodicity, amount of reports, etc, and other associated parameters, (e.g., a time window for output / prediction accuracy verification, measurement accuracy bounds or thresholds for switching report configurations, etc.) Other AI / ML specific hyper parameters and time steps may also be transferred to UE 202 at this step. This message may be an RRC message. At step 218, UE 202 performs measurement procedures (which may be LI and / or L3 measurement procedures) and collects measurement data for determining an output (in some examples determining an output may comprise using AI / ML functionalities to determine said output). At step 220, UE 202 determines an output. This output may be determined using the measurement data collected at step 218, and additionally or alternatively using UE 202 measurement data more broadly. The output may be determined at least partly based on an ML solution. UE 202 may perform AI / ML based functionalities by using measurement data as the input to generate the output. In UE-sided embodiments, AI / ML LCM procedures (e.g., training, inference, monitoring, etc.) may be conducted at the UE side. In one example use case, a time-series long short-term memory (LSTM) prediction model can be used to predict a sequence of RRM measurements or measurement events for mobility. An example method of determining an output (RRM measurement predictions for L3 mobility in the example) using an AI / ML model, including model layout, input and output structure is depicted in more detail later in the specification. At step 222 UE 202 verifies the accuracy verification conditions. This may be performed within a network configured evaluation time window (which may be a time window configured at step 216). Verifying the accuracy verification conditions may comprise determining an accuracy of the output, and comparing this to accuracy verification conditions. The accuracy verification conditions may comprise a threshold accuracy, and UE 202 may switch to a periodic reporting configuration (which may have been configured at step 216) if predicted measurement accuracy level is relatively low (compared to the threshold). For example, if the periodic reporting configuration is expected to provide a higher accuracy than the current reporting configuration (for example, because the periodic reporting configuration comprises more frequent or more accurate measurements), which may be a legacy or default reporting configuration, the periodic reporting configuration may be configured for use below the accuracy threshold. In another example, if the periodic reporting configuration is expected to provide a lower accuracy than the current reporting configuration (for example, because the periodic reporting configuration comprises less frequent or less accurate measurements, which may be less resource intensive), which may be a legacy or default reporting configuration, the periodic reporting configuration may be configured for use above the accuracy threshold. There are multiple potential ways of formulating an accuracy-based verification condition. Three example formulations of accuracy, which may for example be compared to thresholds in an accuracy verification step, are expressed below. The evaluation window size is denoted as Leval, the verification data set collected within this window is Deval. The verification data set may comprise one or more outputs of step 220. For example, the verification data set may comprise a set of predictions of measurements. Option 1: The accuracy fE of the output (compared to a ground truth) may be expressed as a percentage or fraction of the verification data set that match the ground truth. This may be calculated using the equation fE = —-—Xied where y, is data of the lDevail eval output and yt is the ground truth for the i-th sample of the verification data. 1( ) is the indicator function. In the case that the output is a prediction of a measurement or another quantity that may be derived from one or more measurement, the ground truth may be that measurement or quantity (i.e., the ground truth may be determined based on hindsight). Option 2: The accuracy fE of the output may be expressed as a Mean Square Error (where the errors are the differences between an output and a corresponding ground truth). This may be calculated using the equation fE = —-—Xied SQyt - yd}2, where yt is the prediction I Deva! I eval and yt is the ground truth for the t-th sample. Option 3: The accuracy fE of the output may be expressed as a Negative log likelihood. In some examples the output may comprise a confidence score assigned to a prediction or determination (for example, a confidence score could be assigned to a beam or cell, if the output is a beam or cell, or a predicted beam or cell). The Negative log likelihood may be calculated from the negative log of the confidence scores assigned to the prediction or determination that matches a corresponding ground truth (which may not necessarily be outputs having the highest score). This may be calculated using the equation fE = ;SieDevo(l°g(Pi x Pr(51 = yd), where pt is the confidence score of the i-th prediction or determination. As an example, the exact probabilistic value for p( can be obtained from the output of a softmax activation function in the output layer of an ML output. Pr(y, = y() denotes the model accuracy. In this example, in which outputs (which may be ML predictions) are determined at UE 202 ground truth / correct labels may be available at the UE side for each output yt. This information can be used for accuracy evaluation when this procedure is executed at the UE side. There are several methods of determining a ground truth for use in an accuracy verification step. The ground truth may be based on information collected after the data upon which the verification data set was determined was collected (i.e., based on hindsight). One example of this would be using a later measured value as the ground truth for a prediction of that measured value. Additionally or alternatively a ground truth may be determined using a larger data set than the data set used to determine the verification data (e.g., data that is collected more frequently or that spans a larger time than the data used to determine verification data). Additionally or alternatively a ground truth may be determined using a more computationally intensive method than the method used to determine the verification data (e.g., using a method that could not normally be carried out in the time allowed for the determination of the output normally). Additionally or alternatively a ground truth may be determined using more accurate data than the data set used to determine the verification data (e.g., collecting more accurate data may require additional resources). In other examples, the ground truth may be obtained from an external source. For example, the data set from which the verification data is determined may be determined by a testing / verification setup, and the ground truth may correspond to a "correct" output based on that testing / verification setup, provided for the purposes of verifying accuracy. In an example that implements the above mentioned accuracy-based verification conditions, the NW (or NW node 204 of the NW) configures a threshold £th that represents an acceptable prediction accuracy error. If / e <£th / (i e., fE <£th is true) then the verification phase Is set to be qualified and UE shall perform the new periodic reporting framework. If / e >£th, 0-e., Ze <£th is false) then the verification phase is set to be NOT qualified, and UE shall NOT perform the new periodic reporting framework. The legacy reporting framework with high fidelity report will still be used. In this example the new reporting framework may comprise a lower fidelity report than the legacy, default, or current reporting framework provides. At step 224 UE 202 sends a request or indication to network node 204 to switch to the periodic reporting procedures. This request may be a request to switch to the periodic reporting procedures once the accuracy conditions are verified at step 222 (i.e., the accuracy conditions may not be verified at this stage, and the switch may be conditioned on verification). In some embodiments this message may simply convey the evaluation outcome that can be encoded in a bitmap. In some embodiments this message may comprise 1 bit information. Based on a verification of an accuracy against a threshold, UE 202 may send a message 1 or 0 to indicate whether an accuracy verification condition is true or not. In some embodiments this message may comprise n-bit Information. When there are multiple accuracy thresholds (which may correspond to multiple accuracy conditions and multiple potential reporting configurations), a generalized n-blt sequence indication may be sent from UE 202. The indication message of step 224 may be sent over LI or L2 signalling channels, e.g., via MAC control elements (MAC CE) or via uplink control information (UCI). At step 226, network node 204 may evaluate the message of step 224 and prepare the corresponding UE 202 preferred reporting configuration accordingly. In some embodiments this step may be omitted. At step 228, network node 204 sends a response to UE 202, indicating that UE 202 Is to switch to the periodic reporting procedures as per the request in step 224. If the request was a request to switch to the periodic reporting procedures once the accuracy conditions are verified, the indication to switch may be conditioned on the accuracy conditions being verified. The UE 202 preferred reporting configuration including periodicity is granted. Network node 204 may send 1 bit information to grant the UE 202 request to perform the periodic reporting as per the request or deny the request. Similarly, L1 / L2 signalling channels may be used for this indication message, e.g., MAC CE or DCI. At step 230 UE 202 applies the new reporting configuration and sends the measurement report (which may have different or relatively lower accuracy) to network node 204. At step 232, network node 204 may buffer the reporting configuration used by the UE. This may enable network node 204 to assess whether the current approach is beneficial to continue or revert to the legacy configurations. In some embodiments this step may be omitted. Figure 3 is a graph 300 illustrating at a high level how accuracy-based event entering and exit conditions may be used in event triggered reporting. As a prediction accuracy decreases overtime, the prediction accuracy may drop below an event entering threshold 310. An accuracy verification step may determine that the accuracy has dropped below the event entering threshold 310, and trigger event triggered reporting 312. Once the accuracy increases above an event exiting threshold 314, an accuracy verification step may determine that the accuracy has risen above the exiting threshold 314, and stop the event triggered reporting 312. Figure 4 shows message flow sequence 400. Message flow sequence 400 takes place between UE 202 and network node 204. Network node 204 may in some embodiments be a gNB. Message flow sequence 400 corresponds to an embodiment in which an algorithm that determines an output is executed at the UE side. The algorithm may be an ML algorithm. The accuracy of the output is determined and verified against an accuracy condition. On the basis of this verification, event based measurement reporting may be triggered. At step 410, network node 204 sends a capability enquiry to UE 202 for performing the dynamic measurement reporting procedure. The enquiry may be sent via RRC signalling. At step 412, similar to step 212, UE 202 sends capability information to network node 204. This capability information may be sent alongside measurement accuracy requirements for an output determined at UE 202. This message may be sent via RRC signalling. The supported reporting frameworks and the corresponding parameters, such as supported report method types, report periodicities, amount of reports, etc., may be included in this message as capability information. At step 414, similar to step 214, network node 204 prepares configuration parameters according to the capability information to create a reporting framework. This reporting framework may be prepared for use in cases where UE 202 uses AI / ML to determine an output, such as RRM measurement prediction (e.g., for L3 mobility). At step 416, similar to step 216, network node 204 sends a (re)configuration message to UE 202 with the configuration parameters to support the new reporting framework e.g., type of report, periodicity, amount of reports, etc, and other associated parameters, (e.g., a time window for output / prediction accuracy verification, measurement accuracy bounds or thresholds for switching report configurations, etc.) Other AI / ML specific hyper parameters and time steps may also be transferred to UE 202 at this step. This message may be an RRC message. In this case, the reporting framework may comprise a reporting configuration that includes accuracy-based event triggered measurement reports. These events may be defined in the reporting configuration, and may include an event start condition, event exit condition, and other associated configuration parameters. For example, two corresponding thresholds may be configured to control the event start and event exit. At step 418, similar to step 218, UE 202 performs measurement procedures (which may be LI and / or L3 measurement procedures) and collects measurement data for determining an output (in some examples determining an output may comprise using AI / ML functionalities to determine said output). At step 420, similar to step 220, UE 202 determines an output. This output may be determined using the measurement data collected at step 218, and additionally or alternatively using UE 202 measurement data more broadly. The output may be determined at least partly based on an ML solution. UE 202 may perform AI / ML based functionalities by using measurement data as the input to generate the output. In UE-sided embodiments, AI / ML LCM procedures (e.g., training, inference, monitoring, etc.) may be conducted at the UE side. In one example use case, a time-series long short-term memory (LSTM) prediction model can be used to predict a sequence of RRM measurements or measurement events for mobility. An example method of determining an output (RRM measurement predictions for L3 mobility in the example) using an AI / ML model, including model layout, input and output structure is depicted in more detail later in the specification. At step 422 UE 202 verifies the accuracy verification conditions. This may be performed within a network configured evaluation time window (which may be a time window configured at step 416). Verifying the accuracy verification conditions may comprise determining an accuracy of the output, and comparing this to accuracy verification conditions. If an accuracy condition is verified, this may trigger measurement reporting (where the accuracy condition corresponds to an event entering trigger), or may end measurement reporting (where the accuracy condition corresponds to an event exiting trigger). For accuracy-based events, the accuracy verification methods and conditions may correspond to those described at step 222 of method 200. For example, two thresholds may be defined, £Enter and eExit, to control the event entering and exit condition based on the accuracy measure. As discussed in connection with step 222, the accuracy may in some examples be one or more of a percentage, a mean squared error, and a negative log likelihood. UE radio quality based events, Mobility KPI based events, and LCM based events may also be defined and used as triggers for event based reporting. UE radio quality based events may use a radio coupling gain based measure as a trigger (e.g., an RSRP difference between the serving cell and strongest neighbor cell may be compared to thresholds RSRPdiff^enter and RSRPdiff^exit for event entering and exit triggers respectively). The RSRP difference may be measured in dBm. Mobility KPI based events may use mobility KPI related counters as triggers. For example, one or more of a number of handovers (HOs), a number of handover failures (HOFs), a number of Ping-Pongs (PPs), a number of unnecessary HOs, and a number of Radio Link Failures (RLFs), or some measure based on one or more of these measures, may be compared to thresholds MobCounter_enter and MobCounter_exlt. LCM based events may be constructed based on different scenarios behind LCM procedures. For example, a data collection event for model training may be defined, and may trigger high-fidelity measurement reporting. A model monitoring event may also be defined to verify the accuracy of the model if performance degradation is detected, and lower accuracy measurement reporting may be triggered accordingly. At step 424 UE 202 sends an event triggered measurement report to network node 204 when the event entering condition is verified. Using different types of events, UE 202 may perform measurement reporting procedures with different accuracy levels, and may meet different accuracy requirements. For example, different kinds of event trigger (e.g., accuracy based and radio quality based), or event triggers associated with different thresholds (e.g., two different accuracy based thresholds), may result in reporting procedures with different accuracy levels being performed. At step 426, similar to step 232, network node 204 may buffer the reporting configuration used by the UE. This may enable network node 204 to assess whether the current approach is beneficial to continue or revert to the legacy configurations. In some embodiments this step may be omitted. Figure 5 shows message flow sequence 500. Message flow sequence 500 takes place between UE 202 and network node 204. Network node 204 may In some embodiments be a gNB. Message flow sequence 500 corresponds to an embodiment in which an algorithm that determines an output is executed at the UE side. The algorithm may be an ML algorithm. The accuracy of the output is determined and verified against an accuracy condition. On the basis of this verification, the reporting configuration used by UE 202 may be changed. Message flow sequence 500 depicts changing the reporting configuration to a reporting configuration having a semi-persistent reporting type. Accuracy condition verification may be repeated to determine further reporting configurations / periods. Steps 510 - 522 correspond to steps 210 - 222 of method 200, except that in this case the reporting configuration is a semi-persistent reporting configuration, rather than a periodic reporting configuration. At step 524, UE 202 sends a request or Indication to network node 204 to switch to the semi-persistent reporting procedures. This request may be a request to switch to the semi-persistent reporting procedures once the accuracy conditions are verified at step 522 (l.e., the accuracy conditions may not be verified at this stage, and the switch may be conditioned on verification). In some embodiments this message may simply convey the evaluation outcome that can be encoded in a bitmap. In some embodiments this message may comprise 1 bit information. Based on a verification of an accuracy against a threshold, UE 202 may send a message 1 or 0 to indicate whether an accuracy verification condition is true or not. In some embodiments this message may comprise n-bit information. When there are multiple accuracy thresholds (which may correspond to multiple accuracy conditions and multiple potential reporting configurations), a generalized n-bit sequence indication may be sent from UE 202. The indication message of step 524 may be sent over LI or L2 signalling channels, e.g., via MAC control elements (MAC CE) or UCI. At step 526, network node 204 may evaluate the message of step 524 and prepare the corresponding UE 202 preferred reporting configuration accordingly. In some embodiments this step may be omitted. At step 528, network node 204 sends a response to UE 202, Indicating that UE 202 Is to switch to the semi-persistent reporting procedures as per the request in step 524. If the request was a request to switch to the semi-persistent reporting procedures once the accuracy conditions are verified, the indication to switch may be conditioned on the accuracy conditions being verified. The UE 202 preferred reporting configuration including periodicity is granted. Additionally, network node 204 may also configure the UE with a method of switching the reporting periodicity. Network node 204 may send 1 bit information to grant the UE 202 request to perform the semi-persistent reporting as per the request or deny the request. Similarly, L1 / L2 signalling channels may be used for this indication message, e.g., MAC CE or DCI. At step 530, UE 202 applies the new reporting configuration and sends the measurement report to network node 204. The report may have a configured accuracy level, and the reporting period is Ti. Reports are therefore sent every Ti. At step 532, UE 202 monitors the accuracy based verification conditions. Once these conditions no longer hold true, UE 202 may switch to a different reporting period, T2. At step 534, UE applies a new reporting configuration, which may have a different reporting period and / or a different accuracy level to the reports of step 530 to network node 204. The reporting period at step 534 may be T2, so in some examples T2 * Ti. In above described embodiments, measurement reporting is changed dynamically based on whether outputs of an algorithm or process (which may be AI / ML algorithms or processes) maintained at the UE side meet accuracy conditions. In further embodiments, measurement reporting may be dynamic based on whether outputs of an algorithm or process maintained at the network side meets accuracy conditions. This may be additional to or alternative to dynamic reporting based on the accuracy of a UE side algorithm or process. Figure 6 shows message flow sequence 600. Message flow sequence 600 takes place between UE 202 and network node 204. Network node 204 may in some embodiments be a gNB. Message flow sequence 600 corresponds to an embodiment in which an algorithm that determines an output is executed at the network side. The algorithm may be an ML algorithm. The accuracy of the output is determined and verified against an accuracy condition. On the basis of this verification, the reporting configuration used by UE 202 may be changed. Message flow sequence 600 depicts changing the reporting configuration to a reporting configuration having a periodic reporting type. Steps corresponding to 210 and 212 may be included in this process (e.g., before step 610). These steps are not essential, as the network may rely on capabilities that UE 202 can be assumed to possess, and UE 202 may not have accuracy requirements of its own to communicate. Message flow sequence 600 starts at step 610, which is similar to to step 214. At step 610 network node 204 prepares configuration parameters to create a reporting framework. This reporting framework may be prepared for use in cases where network 204 uses AI / ML to determine an output, such as RRM measurement prediction (e.g., for L3 mobility). At step 612, network node 204 sends a (re)configuration message to UE 202 with the configuration parameters to support the new reporting framework e.g., type of report, periodicity, amount of reports, etc. This message may be an RRC message. At step 614, UE 202 performs measurement procedures (which may be LI and / or L3 measurement procedures) and collects measurement data for use by network node 204 in determining an output. At step 616, UE 202 sends a measurement report to network node 204. This measurement report may comprise data collected at step 614. As the network may not have made a decision to use a new (i.e., a non-default or non-legacy) reporting configuration yet, this measurement report may be a default or legacy LI or L3 measurement report. At step 618, network node 204 determines an output. This output may be determined using (i.e., based at least in part on) the measurement data from the measurement report received at step 616. The output may be determined at least partly based on an ML solution. UE 202 may perform AI / ML based functionalities by using UE measurement data as the input to generate the output. In NW-sided embodiments, AI / ML LCM procedures (e.g., training, inference, monitoring, etc.) may be conducted at the network side. In one example use case, a time-series long short-term memory (LSTM) prediction model can be used to predict a sequence of RRM measurements or measurement events for mobility. An example method of determining an output (RRM measurement predictions for L3 mobility in the example) using an AI / ML model, including model layout, input and output structure is depicted in more detail later in the specification. At step 620, network node 204 verifies the accuracy verification conditions. This may be performed within an evaluation time window, which may be network configured. Verifying the accuracy verification conditions may comprise determining an accuracy of the output, and comparing this to accuracy verification conditions. The accuracy verification process carried out at the network side at step 620 may be similar to the accuracy verification process carried out at the UE side in step 222 of message flow sequence 200, but carried out at the network side, based on outputs calculated at the network. At step 622, network node 204 may determine a preferred reporting period for measurement reports by UE 202. This preferred reporting period may in some examples be determined at an earlier stage (it may for example be sent in the RRC configuration message of step 612). At step 624, network node 204 sends an indication to UE 202 that it is to switch measurement report configuration (in this case, to a measurement report configuration having periodic reporting procedures). The indication may include an indication of reporting period. L1 / L2 signalling channels may be used for this indication message, e.g., MAC CE or DCI. At step 626a - 626c, UE 202 applies the new reporting configuration and sends measurement reports (which may have different or relatively lower accuracy to a legacy or default measurement report) to network node 204. At step 628, network node 204 may buffer the reporting configuration used by the UE. This may enable network node 204 to assess whether the current approach is beneficial to continue or revert to the legacy configurations. In some embodiments this step may be omitted. Figure 7 shows message flow sequence 700. Message flow sequence 700 takes place between UE 202 and network node 204. Network node 204 may in some embodiments be a gNB. Message flow sequence 700 corresponds to an embodiment in which an algorithm that determines an output is executed at the network side. The algorithm may be an ML algorithm. The accuracy of the output is determined and verified against an accuracy condition. On the basis of this verification, the reporting configuration used by UE 202 may be changed. Message flow sequence 600 depicts changing the reporting configuration to a reporting configuration having an event triggered reporting type. Steps corresponding to 210 and 212 may be included in this process (e.g., before step 710). These steps are not essential, as the network may rely on capabilities that UE 202 can be assumed to possess, and UE 202 may not have accuracy requirements of its own to communicate. Message flow sequence 700 starts at step 710. Steps 710 - 720 may correspond to steps 610 - 620, however, the configuration parameters at step 610 may correspond to a reporting configuration having an event triggered type. A step corresponding to step 622 may be unnecessary in this case, as the reporting is event triggered rather than periodic. At step 722, network node 204 sends an indication to UE 202 that it is to switch measurement report configuration (in this case, to a measurement report configuration having event triggered reporting). L1 / L2 signalling channels may be used for this indication message, e.g., MAC CE or DCI. At step 724, UE 202 performs measurement procedures (which may be LI and / or L3 measurement procedures) and collects measurement data for use by network node 204. At step 726, UE 202 sends measurement reports to network node 204 according to the configured event based reporting configuration. Figure 8 shows message flow sequence 800. Message flow sequence 800 takes place between UE 202 and network node 204. Network node 204 may in some embodiments be a gNB. Message flow sequence 800 corresponds to an embodiment In which an algorithm that determines an output is executed at the network side. The algorithm may be an ML algorithm. The accuracy of the output is determined and verified against an accuracy condition. On the basis of this verification, the reporting configuration used by UE 202 may be changed. Message flow sequence 800 depicts changing the reporting configuration to a reporting configuration having a semi-persistent reporting type. Steps corresponding to 210 and 212 may be included in this process (e.g., before step 810). These steps are not essential, as the network may rely on capabilities that UE 202 can be assumed to possess, and UE 202 may not have accuracy requirements of its own to communicate. Message flow sequence 800 starts at step 810. Steps 810 - 820 may correspond to steps 610 - 620, however, the configuration parameters at step 810 may correspond to a reporting configuration having a semi-persistent reporting type. At step 822, network node 204 sends an indication to UE 202 that it is to switch measurement report configuration (in this case, to a measurement report configuration having periodic reporting procedures). The indication may indicate a reporting period Ti. L1 / L2 signalling channels may be used for this indication message, e.g., MAC CE or DCI. At step 824a - 824c, UE 202 applies the new reporting configuration and sends measurement reports (which may have different or relatively lower accuracy to a legacy or default measurement report) to network node 204. The reporting period is Ti. Reports are therefore sent every Ti. At step 826, network node 204 monitors the accuracy based verification conditions. At step 828, once the accuracy verification conditions determined at step 820 are determined at step 826 to no longer hold true, network node 204 may indicate to UE 202 that it is to switch to a different reporting period, T2. L1 / L2 signalling channels may be used for this indication message, e.g., MAC CE or DCI. At step 830a - 830c, UE 202 applies a new reporting configuration, which may have a different reporting period and / or a different accuracy level to the reports of step 530 to network node 204. The reporting period at step 534 may be T2, so in some examples T2 * Ti. Figure 9 is a block diagram showing an example apparatus 900 in accordance with example embodiments of the invention. Apparatus 900 comprises a processor 910 and memory 912. Processor 910 may execute instructions stored in memory 912. Memory 912 may store instructions for execution by processor 910. In some embodiments, apparatus 900 may be a user device, such as UE 202. In some embodiments, apparatus may be a network node, such as network node 204. Network node and apparatus 900 may in some embodiments be a gNB. Processor 910 may be coupled to a receiver and / or transmitter, which may allow apparatus 900 to respectively receive and / or send data to other apparatuses. Processor 910 may optionally be coupled to measurement device 914. For example a receiving antenna may be considered a measurement device capable of obtaining radio signals. From radio signals further measurements may be derived through various well known measurement devices and processing techniques to determine quantities such as RSRP. Figure 10 is a flow diagram showing an example method 1000 in accordance with example embodiments of the invention. Method 1000 may be executed by apparatus 900 in some example embodiments. Method 1000 starts at step 1010. At step 1010 a reporting framework is obtained. The reporting framework maps one or more accuracy conditions to one or more reporting behaviours and / or one or more changes of reporting behaviour. The one or more accuracy conditions may comprise at least one or more of the following: a threshold accuracy above which an accuracy condition is met; a threshold accuracy below which an accuracy condition is met; an accuracy range within which an accuracy condition is met; and an accuracy range outside of which an accuracy condition is met. One or more accuracies of the one or more accuracy conditions may be defined in terms of at least one or more of: a proportion of one or more outputs that match a one or more corresponding ground truths; a mean square error of one or more outputs with respect to one or more corresponding ground truths; and one or more negative logarithms of one or more likelihoods, wherein the likelihoods are likelihoods assigned to outputs that match corresponding ground truths. Based upon the selected measure of accuracy, the accuracy conditions may be expressed correspondingly. Different expressions of accuracy may be appropriate for different outputs and methods of determining outputs. For example, mean squared error may be less appropriate when the output is an ID, but may be more appropriate when the output is a measurement or predicted measurement of a value defined on a continuous scale, such as R.SRP. At step 1012, an indication of whether an output meets an accuracy condition of the one or more accuracy conditions is obtained. The output may comprise one or more of: a predicted value of a measurement; a target cell; and a target beam. At step 1014, a reporting behaviour and / or one or more changes of reporting behaviour is determined based at least in part upon the obtained reporting framework and the obtained indication of whether an accuracy condition of the one or more accuracy conditions is met by the output. The one or more reporting behaviours may each comprise the use of a respective measurement report configuration. This may permit dynamic changing of measurement report configuration based on accuracy conditions, and therefore provide dynamic reporting behaviour based on accuracy conditions. The respective measurement report configurations of the one or more reporting behaviours may each comprise at least one or more of at least the following parameters: a reporting criteria; an event that is to trigger reporting; a report period / interval; and an amount of reports. The respective measurement report configurations of the one or more reporting behaviours may comprise at least one or more accuracy levels at which one or more user device measurements are to be collected. The reporting framework may comprise first and second reporting behaviours. A first reporting behaviour of the one or more reporting behaviours may comprise a first measurement report configuration comprising a first accuracy level at which one or more user device measurements are to be collected. A second reporting behaviour of the one or more reporting behaviours may comprise a second measurement report configuration comprising a second accuracy level at which one or more user device measurements are to be collected, wherein the first accuracy level is higher than the second accuracy level. A first accuracy condition may be met above a threshold and map to the second reporting behaviour and a second accuracy condition may be met met below a threshold and map to the first reporting behaviour. This allows dynamic switching between measurement report configurations having different accuracy. In some example embodiments the reporting behaviours may comprise event triggered measurement reports. The reporting framework may comprise one or more measurement report configurations mapping the one or more accuracy conditions to triggers of event triggered measurement reports. Using accuracy conditions to trigger event triggered measurement reports, measurement reports may be sent dynamically in response to accuracy conditions. In some embodiments method 1000 is carried out by a user device (such as UE 202), and further comprises collecting measurement data associated with radio access network signals (using measurement device 914 for example). Obtaining the reporting framework may comprise receiving the reporting framework from a network node (such as network node 204), and obtaining the indication of whether an output meets an accuracy condition may comprise generating the output based on the obtained measurement data. An indication of the accuracy of the generated output may be determined based at least in part on the generated output, and whether the generated output meets the accuracy condition may be verified based on the indication of the accuracy and the accuracy condition. If method 1000 is carried out by a user device, method 1000 may further comprise, in response to obtaining an indication that the output meets an accuracy condition, sending a request to a network node to use a measurement report configuration comprised by the reporting behaviour mapped to that accuracy condition by the reporting framework. The user device may then receive from the network node an indication that the measurement report configuration is to be used by user device, and apply the measurement report configuration. If method 1000 is carried out by a user device, and the accuracy condition met by the output is mapped to a reporting behaviour comprising a measurement report configuration having an event-based reporting criteria, the method may further comprise verifying that an event identified in the measurement configuration has occurred and triggering a measurement report based on verifying that the event has occurred. If method 1000 is carried out by a user device, the method may further comprise sending to a network node capability information associated with the user device and accuracy requirements of the outputs of the user device. The network node may provide a more suitable reporting framework if it is based on the capabilities and requirements of the user device. In some embodiments, method 1000 may be carried out by a network node (e.g., apparatus 900 may be a network node, such as a gNB). Where method 1000 is carried out by a network node, it may further comprise receiving measurement data associated with radio access network signals from a user device, and obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions may comprise generating the output based on the measurement data received from the user device, determining an accuracy of the generated output, and verifying whether the generated output meets the accuracy condition. The network node may therefore verify the accuracy of outputs of a network side process that uses user device measurement data as an input. Where method 1000 is carried out by a network node, the method may further comprise sending to the user device an instruction to use a measurement report configuration comprised by the reporting behaviour mapped to the met accuracy condition. The network node may therefore dynamically change measurement reporting based on an accuracy of an output of a network side process. Where method 1000 is carried out by a network node, the method may further comprise receiving from the user device capability information, and obtaining a reporting framework may comprise generating a reporting framework based on capability information associated with a user device and accuracy requirements for the output. In some examples a reporting framework may rely on user device capabilities that a user device can be expected to have, so it may not be necessary to obtain capability information from the user device in all examples. In some embodiments, method 1000 may be carried out by a network node, and obtaining a reporting framework may comprise generating a reporting framework based on capability Information associated with a user device and accuracy requirements of outputs of a user device. The method may further comprise sending the one or more accuracy conditions to the user device and receiving from a user device an indication of whether an output meets an accuracy condition of the one or more accuracy conditions. The apparatus may then send to the user device an instruction to use a measurement report configuration comprised by the reporting behaviour mapped to the met accuracy condition. Therefore, the output and whether the output meets an accuracy verification condition may be determined at a user device and indicated to the network, and the network may decide on the basis of whether the accuracy verification condition is met to instruct the user device to use a measurement report configuration, providing dynamic measurement reporting. Figure 11 is a flow diagram showing an example method 1100 in accordance with example embodiments of the invention. Method 1100 may be executed by apparatus 900 in some example embodiments. Method 1100 starts at step 1110. At step 1010, user device measurement data is obtained. In some embodiments user device measurement data is obtained from measurement device 914 coupled to apparatus 900, and apparatus 900 is a user device. In some embodiments, user device measurement data is obtained from a further apparatus. For example, user device measurement data may be received at apparatus 900 from a user device. User device measurement data may comprise measurements associated with radio access network signals. User device measurement data may comprise measurement data indicating one or more of: a reference signal received power; a reference signal received quality; a received signal strength indicator; a signal to interference plus noise ratio; and a signal to noise ratio. User device measurement data may be collected at different levels of accuracy (for example, using different hardware or combinations of hardware that produce measurements of different accuracies). At step 1112, user device measurement data is processed to determine an output. Processing the user device measurement data may comprise using AI / ML models to determine an output. The output may predict measurement data values. The output may predict a target cell or a target beam. At step 1114, an indication of the accuracy of the output is determined. The accuracy may be determined based at least in part on the output and a ground truth. The ground truth may be determined based on at least one of: second measurement data that is more accurate than the user device measurement data; a processing method that is more computationally intensive than a processing method used in the step of processing user device measurement data to determine the output; third measurement data that comprises more data than user device measurement data; and fourth measurement data that comprises data collected after the user device measurement data was collected. At step 1116, the indicated accuracy is compared with one or more accuracy conditions. The indication of an accuracy of the output and the one or more accuracy conditions may be defined in terms of at least one or more of: a proportion of determined outputs that match a ground truth; a mean square error of determined outputs compared to a ground truth; and a negative log likelihood. The accuracy conditions may comprise one or more of: a threshold accuracy above which an accuracy condition is met, a threshold accuracy below which an accuracy condition is met, an accuracy range within which an accuracy condition is met, and an accuracy range without which an accuracy condition is met. ML model Implementation example Figure 12 is a diagram showing schematically an implementation example of a ML timeseries prediction model 1200. Model 1200 may be used for inter-frequency mobility management and HO decision making. In the example structure of model 1200, the input frame 1210 can be a sequence of L2-RSRP values 1212 and / or a sequence of beam IDs 1214. These two inputs may be concatenated in operation 1220 before feeding into an ML model. This example uses L2-RSRP input values from 294 beams (14 beams per gNB from 21 gNBs) within a 3 second window. In this example, the gNBs measure synchronisation signal blocks (SSBs) every 20ms, so the measurement duration covers 150 samples if the total input length is 3 seconds. The input frame, (i.e., L2-RSRP + Beam IDs) are fed into the ML model, which in this example is a long-short-term-memory (LSTM) recurrent neural network (RNN) to obtain time sequence output 1260. In this example the concatenated input is fed into one or more dense layers 1230 before being input to one or more LSTM / RNN layers 1240, before being input to another one or more dense layers 1250. The output 1260 of the model can be any of predicted L2-RSRP values, or key markers, e.g., HO indicator metric / predicted HO events within the prediction window. The main benefit of model 1200 is that it allows the delivery of additional information in terms of beam / cell selection on top of the legacy L3 mobility management framework. Input As described above, the input sequence at time t is denoted as (Qt-i« -.Qi-n)- Each sample Q is a tuple of: Q = {r,b} = {{r^,... b2.....6294}} r = {ri,r2, -.^94} is the input RSRP vector measured from from 294 beams (14 beams from 21 gNBs), i.e., b = {b1,b2,...,b294}. The Input frame is measured from a past N ms that contains q past samples for each measurement or beam ID. In this example, a gNB measures SSB every 20 ms, q = 150 measurement samples are collected within the total input length N = 3000 ms. So the input data is thus a 150 x 588 matrix. Output (Labe!) The labeled data used to train the model is the sequence of RSRP samples in a given prediction window M ms that contains m past samples for each measurement or beam ID. In this example, the prediction output is the RSRP measured from from 294 beams (14 beams from 21 gNBs) in M = 100 ms that include m = 5 samples for each index. The label data is thus a 294 x 5 matrix of (Pt+1, ...Pt+M). Loss function A time series prediction model is implemented to learn the dependencies of historical RSRP plus each beam index (Q^1( ...,Q£^A) over the future RSRP values (Pt+1, ...Pt+U). LSTM has been proven as an effective approach to tackle the long-range dependencies problem, which is used in our model design in Fig. 10. The present model comprises a set of parameters associated with each input Qt. The earlier 0(W are mutlipled with a weight and added to the later one, to capture the time series dependencies. The output layer is associated with each label P^ A loss function is defined to evaluate the error between the model predicted P and detected P, for example: k e(P,P)=^(jt(P\Q,6)-Pt)2 t=i Training In the training process, optimization algorithm like Stochastic Gradient Descent (SGD) can be applied to tune each 0(u), such that the average prediction error of samples collected from the UE can be minimized. The model reliability evaluation and assessment are conducted when training steps are completed. ML model Implementation example Figure 12 shows schematically an implementation example of a ML time-series prediction model 1200. Model 1200 may be used for inter-frequency mobility management and HO decision making. In the example structure of model 1200, the input frame 1210 can be a sequence of L2-RSRP values 1212 and / or a sequence of beam IDs 1214. These two inputs may be concatenated in operation 1220 before feeding into an ML model. This example uses L2-RSRP input values from 294 beams (14 beams per gNB from 21 gNBs) within a 3 second window. In this example, the gNBs measure synchronisation signal blocks (SSBs) every 20ms, so the measurement duration covers 150 samples If the total input length is 3 seconds. The input frame, (i.e., L2-RSRP + Beam IDs) are fed into the ML model, which in this example Is a long-short-term-memory (LSTM) recurrent neural network (RNN) to obtain time sequence output 1260. In this example the concatenated input is fed into one or more dense layers 1230 before being input to one or more LSTM / RNN layers 1240, before being input to another one or more dense layers 1250. The output 1260 of the model can be any of predicted L2-RSRP values, or key markers, e.g., HO indicator metric / predicted HO events within the prediction window. The main benefit of model 1200 is that it allows the delivery of additional information in terms of beam / cell selection on top of the legacy L3 mobility management framework. Input As described above, the input sequence at time t is denoted as (Qt_1; -.Qi-n)- Each sample Q is a tuple of: Q = {r,b} = {{^,^,...,^94),(^, / ,2.....b294}} r = (r1(r2,...,r294) is the input RSRP vector measured from from 294 beams (14 beams from 21 gNBs), i.e., b = {b^, ...,b2g4}. The input frame is measured from a past N ms that contains q past samples for each measurement or beam ID. In this example, a gNB measures SSB every 20 ms, q = 150 measurement samples are collected within the total input length N = 3000 ms. So the input data is thus a 150 x 588 matrix. Output (Labe!) The labeled data used to train the model is the sequence of RSRP samples in a given prediction window M ms that contains m past samples for each measurement or beam ID. In this example, the prediction output Is the RSRP measured from from 294 beams (14 beams from 21 gNBs) in M = 100 ms that Include m = 5 samples for each index. The label data is thus a 294 x 5 matrix of (Pt+1,... P{+M)• Loss function k time series prediction model is implemented to learn the dependencies of historical RSRP plus each beam index (0^,...,0^) over the future RSRP values ...Pt+U). LSTM has been proven as an effective approach to tackle the long-range dependencies problem, which is used in our model design in Fig. 10. The present model comprises a set of parameters e(in, associated with each input Qt. The earlier 0(W are mutlipled with a weight and added to the later one, to capture the time series dependencies. The output layer 0(u) is associated with each label Pi. A loss function is defined to evaluate the error between the model predicted P and detected P, for example: k i=l Training In the training process, optimization algorithm like Stochastic Gradient Descent (SGD) can be applied to tune each such that the average prediction error of samples collected from the UE can be minimized. The model reliability evaluation and assessment are conducted when training steps are completed. For completeness, Figure 13 is a schematic diagram of components of one or more of the example embodiments described previously, which hereafter are referred to generically as a processing system 1900. The processing system 1900 may, for example, be comprised by the device referred to in the claims below. The processing system 1900 may have a processor 1902, a memory 1904 closely coupled to the processor and comprised of a RAM 1914 and a ROM 1912, and, optionally, a user input 1910 and a display 1918. The processing system 1900 may comprise one or more network / apparatus interfaces 1908 for connection to a network / apparatus, e.g. a modem which may be wired or wireless. The network / apparatus interface 1908 may also operate as a connection to other apparatus such as device / apparatus which is not network side apparatus. Thus, direct connection between devices / apparatus without network participation is possible. The processor 1902 is connected to each of the other components in order to control operation thereof. The memory 1904 may comprise a non-volatile memory, such as a hard disk drive (HDD) or a solid state drive (SSD). The ROM 1912 of the memory 1904 stores, amongst other things, an operating system 1915 and may store software applications 1916. The RAM 1914 of the memory 1904 is used by the processor 1902 for the temporary storage of data. The operating system 1915 may contain code which, when executed by the processor implements aspects of the methods 1000 and 1100 described above, along with aspects of the message flow sequences 200, 400, 500, 600, 700, and 800. Note that in the case of small device / apparatus the memory can be most suitable for small size usage i.e. not always a hard disk drive (HDD) or a solid state drive (SSD) is used. The processor 1902 may take any suitable form. For instance, it may be a microcontroller, a plurality of microcontrollers, a processor, or a plurality of processors. The processing system 1900 may be a standalone computer, a server, a console, or a network thereof. The processing system 1900 and needed structural parts may be all inside device / apparatus such as loT device / apparatus i.e. embedded to very small size. In some example embodiments, the processing system 1900 may also be associated with external software applications. These may be applications stored on a remote server device / apparatus and may run partly or exclusively on the remote server device / apparatus. These applications may be termed cloud-hosted applications. The processing system 1900 may be in communication with the remote server device / apparatus in order to utilize the software application stored there. FIG. 14 shows a tangible media, in the form of a removable memory unit 2010, storing computer-readable code which when run by a computer may perform methods according to example embodiments described above. The removable memory unit 2010 may be a memory stick, e.g. a USB memory stick, having internal memory 2030 storing the computer-readable code. The internal memory 2030 may be accessed by a computer system via a connector 2020. Of course, other forms of tangible storage media may be used, as will be readily apparent to those of ordinary skilled in the art. Tangible media can be any device / apparatus capable of storing data / information which data / information can be exchanged between devices / apparatus / network. Embodiments of the present invention may be implemented In software, hardware, application logic or a combination of software, hardware and application logic. The software, application logic and / or hardware may reside on memory, or any computer media. In an example embodiment, the application logic, software or an instruction set is maintained on any one of various conventional computer-readable media. In the context of this document, a "memory" or "computer-readable medium" may be any non-transitory media or means that can contain, store, communicate, propagate or transport the instructions for use by or in connection with an instruction execution system, apparatus, or device, such as a computer. Reference to, where relevant, "computer-readable medium", "computer program product", "tangibly embodied computer program" etc., or a "processor" or "processing circuitry" etc. should be understood to encompass not only computers having differing architectures such as single / multi-processor architectures and sequencers / parallel architectures, but also specialised circuits such as field programmable gate arrays FPGA, application specify circuits ASIC, signal processing devices / apparatus and other devices / apparatus. References to computer program, instructions, code etc. should be understood to express software for a programmable processor firmware such as the programmable content of a hardware device / apparatus as instructions for a processor or configured or configuration settings for a fixed function device / apparatus, gate array, programmable logic device / apparatus, etc. If desired, the different functions discussed herein may be performed in a different order and / or concurrently with each other. Furthermore, if desired, one or more of the abovedescribed functions may be optional or may be combined. Similarly, it will also be appreciated that the flow and signalling diagrams of Figures 2, 4, 5, 6, 7, 8 ,10, and 11 are examples only and that various operations depicted therein may be omitted, reordered and / or combined. It will be appreciated that the above-described example embodiments are purely illustrative and are not limiting on the scope of the invention. Other variations and modifications will be apparent to persons skilled in the art upon reading the present specification. Moreover, the disclosure of the present application should be understood to include any novel features or any novel combination of features either explicitly or implicitly disclosed herein or any generalization thereof and during the prosecution of the present application or of any application derived therefrom, new claims may be formulated to cover any such features and / or combination of such features. Although various aspects of the invention are set out in the independent claims, other aspects of the invention comprise other combinations of features from the described example embodiments and / or the dependent claims with the features of the independent claims, and not solely the combinations explicitly set out in the claims. 5 It is also noted herein that while the above describes various examples, these descriptions should not be viewed in a limiting sense. Rather, there are several variations and modifications which may be made without departing from the scope of the present invention as defined in the appended claims. io

Claims

1. An apparatus comprising:means for obtaining a reporting framework, wherein the reporting framework maps one or more accuracy conditions to one or more reporting behaviours and / or one or more changes of reporting behaviour;means for obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; andmeans for determining a reporting behaviour and / or one or more changes of reporting behaviour based at least in part upon the reporting framework and the indication of whether an accuracy condition of the one or more accuracy conditions is met by the output.

2. The apparatus of claim 1, wherein:the one or more accuracy conditions comprise at least one or more of the following: a threshold accuracy above which an accuracy condition is met; a threshold accuracy below which an accuracy condition is met; an accuracy range within which an accuracy condition is met; and an accuracy range outside of which an accuracy condition is met.

3. The apparatus of claim 1 or claim 2, wherein one or more accuracies of the one or more accuracy conditions are defined in terms of at least one or more of:a proportion of one or more outputs that match a one or more corresponding ground truths;a mean square error of one or more outputs with respect to one or more corresponding ground truths; andone or more negative logarithms of one or more likelihoods, wherein the likelihoods are likelihoods assigned to outputs that match corresponding ground truths.

4. The apparatus of any of claims 1-3, wherein the output comprises one or more of:a predicted value of a measurement;a target cell; anda target beam.

5. The apparatus of any of claims 1-4, wherein:the one or more reporting behaviours each comprise the use of a respective measurement report configuration.

6. The apparatus of claim 5, wherein:the respective measurement report configurations of the one or more reporting behaviours each comprise at least one or more of at least the following parameters: a reporting criteria; an event that is to trigger reporting; a report period / interval; and an amount of reports.

7. The apparatus of any of claims 5 or 6, wherein the respective measurement report configurations of the one or more reporting behaviours comprise at least one or more accuracy levels at which one or more user device measurements are to be collected.

8. The apparatus of claim 7, wherein:a first reporting behaviour of the one or more reporting behaviours comprises a first measurement report configuration comprising a first accuracy level at which one or more user device measurements are to be collected;a second reporting behaviour of the one or more reporting behaviours comprises a second measurement report configuration comprising a second accuracy level at which one or more user device measurements are to be collected, wherein the first accuracy level is higher than the second accuracy level;a first accuracy condition is met above a threshold and maps to the second reporting behaviour; anda second accuracy condition is met below a threshold and maps to the first reporting behaviour.

9. The apparatus of any of claims 1-4, wherein:the reporting behaviours comprise event triggered measurement reports; andthe reporting framework comprises one or more measurement report configurations mapping the one or more accuracy conditions to triggers of event triggered measurement reports.

10. The apparatus of any of claims 1 to 9, wherein the apparatus is a user device, and wherein:the apparatus further comprises means for collecting measurement data associated with radio access network signals;the means for obtaining the reporting framework comprises means for receiving the reporting framework from a network node; andthe means for obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions comprises means for generating the output based on measurement data obtained by the user device, means for determining an accuracy of the generated output, and means for verifying whether the generated output meets the accuracy condition.

11. The apparatus of claim 10, wherein the one or more reporting behaviours each comprise the use of a respective measurement report configuration, and the apparatus comprises:means for, responsive to obtaining an indication that the output meets an accuracy condition, sending a request to a network node to use a measurement report configuration comprised by the reporting behaviour mapped to that accuracy condition by the reporting framework;means for receiving from the network node a response indicating that the measurement report configuration is to be used by the apparatus; andmeans for applying the indicated measurement report configuration.

12. The apparatus of claim 11, wherein the accuracy condition met by the output is mapped to a reporting behaviour comprising a measurement report configuration having an event-based reporting criteria; further comprising:means for verifying that an event identified in the measurement configuration has occurred; andmeans for triggering a measurement report based on verifying that the event has occurred.

13. The apparatus of any of claims 10 to 12, further comprising:means for sending to the network node capability information associated with the user device; andmeans for sending to the network node accuracy requirements for the output.

14. The apparatus of any of claims 1 to 9, wherein the apparatus is a network node, and wherein:the apparatus further comprises means for receiving measurement data associated with radio access network signals from a user device;the means for obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions comprises means for generating the output based on the measurement data received from the user device, means fordetermining an accuracy of the generated output, and means for verifying whether the generated output meets the accuracy condition.

15. The apparatus of claim 14, wherein the apparatus comprises means for sending to the user device an instruction to use a measurement report configuration comprised by the reporting behaviour mapped to the met accuracy condition.

16. The apparatus of claim 14 or claim 15 wherein:the apparatus further comprises means for receiving from the user device capability information; andthe means for obtaining a reporting framework comprises means for generating a reporting framework based on capability information associated with a user device and accuracy requirements for the output.

17. The apparatus of any of claims 1 to 9, wherein the apparatus is a network node, and wherein:the means for obtaining a reporting framework comprises means for generating a reporting framework based on capability information associated with a user device and accuracy requirements of outputs of a user device;the apparatus comprises means for sending the one or more accuracy conditions to the user device;the apparatus comprises means for receiving from a user device an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; andthe apparatus comprises means for sending to the user device an instruction to use a measurement report configuration comprised by the reporting behaviour mapped to the met accuracy condition.

18. A method comprising:obtaining a reporting framework, wherein the reporting framework maps one or more accuracy conditions to one or more reporting behaviours and / or one or more changes of reporting behaviour;obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions; anddetermining a reporting behaviour and / or one or more changes of reporting behaviour based at least in part upon the reporting framework and the indication of whether an accuracy condition of the one or more accuracy conditions is met by theoutput.

19. The method of claim 18, wherein:the method further comprises collecting measurement data associated with radio access network signals;obtaining the reporting framework comprises receiving the reporting framework from a network node; andobtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions comprises generating the output based on the collected measurement data, determining an accuracy of the generated output, and verifying whether the generated output meets the accuracy condition.

20. The method of claim 18, wherein:the method further comprises receiving measurement data associated with radio access network signals from a user device;obtaining an indication of whether an output meets an accuracy condition of the one or more accuracy conditions comprises generating the output based on the measurement data received from the user device, determining an accuracy of the generated output, and verifying whether the generated output meets the accuracy condition.

Images