WO2021064275A1

WO2021064275A1 - Radio access information reporting in wireless network

Info

Publication number: WO2021064275A1
Application number: PCT/FI2019/050707
Authority: WO
Inventors: Istvan Kovacs; Mohammad Majid BUTT; Teemu VEIJALAINEN; Jani Moilanen; Wolfgang Zirwas; Tero Henttonen; Luis Guilherme UZEDA GARCIA
Original assignee: Nokia Technologies Oy
Priority date: 2019-10-02
Filing date: 2019-10-02
Publication date: 2021-04-08

Abstract

According to an example aspect of the present invention, there is provided a method comprising,determining at least one parameter of a prediction algorithm; applying an input to the prediction algorithm with the at least one determined parameter to obtain a first inference result for radio access control function; and transmitting to a second wireless device inference conditions information associated with the first inference result and characterizing the conditions under which the first inference result was obtained.

Description

RADIO ACCESS INFORMATION REPORTING IN WIRELESS NETWORK

FIELD

Various example embodiments relate to radio access information reporting in wireless networks, and in particular radio access related inference reporting.

BACKGROUND

In a modem mobile communication systems, a wireless access network node, which may be also be referred to as a base station, determines a transmission format, a transmission block size, a modulation and coding scheme, and the like to be used in a downlink (DL) and an uplink (UL). To perform such determination for the DL, the network node needs information about the performance of a current DL channel from a wireless (user) device, and the information is generally referred to as channel state information (CSI).

Multi-antenna techniques can significantly increase the data rates and reliability of a wireless communication system. The performance is in particular improved if both the transmitter and the receiver are equipped with multiple antennas, which results in a multiple- input multiple-output (MIMO) communication channel.

Wireless devices operating in extremely high frequency (EHF) spectrum, also referred to as the millimetre wave (mmwave) spectrum, are particularly sensitive to sudden blockages. As mmwave signals do not diffract around objects the way in which cellular band signals do, very sudden loss of signal can occur when the line-of-sight (LoS) link between the base station and mobile becomes obstructed by an object, such as a building. This is an example where further improvements can be achieved by proactive radio resource management applying prediction algorithms. SUMMARY OF THE INVENTION

According to some aspects, there is provided the subject-matter of the independent claims. Some embodiments for some or all of the aspects are defined in the dependent claims.

According to a first aspect, there is provided a first method, comprising: determining at least one parameter of a prediction algorithm; applying an input to the prediction algorithm with the at least one determined parameter to obtain a first inference result for radio access control function; and transmitting to a second wireless device inference conditions information associated with the first inference result and characterizing the conditions under which the first inference result was obtained.

According to a second aspect, there is provided a second method, comprising: receiving from a first wireless device inference conditions information associated with a first inference result and characterizing conditions under which the first inference result was obtained by a prediction algorithm in the first wireless device, and generating a second inference result on the basis of the first inference result and the inference conditions information for controlling a radio access function on the basis of the second inference result.

According to a third aspect, there is provided an apparatus, comprising at least one processor, at least one memory including computer program code, the at least one memory and the computer program code being configured to, with the at least one processor, cause the apparatus at least to perform the method of the first aspect or an embodiment thereof.

According to a fourth aspect, there is provided an apparatus, comprising at least one processor, at least one memory including computer program code, the at least one memory and the computer program code being configured to, with the at least one processor, cause the apparatus at least to perform the method of the second aspect or an embodiment thereof

According to a fifth aspect, there is provided a computer program product, a computer readable medium, or a non-transitory computer readable medium comprising program instructions for causing an apparatus to perform the method according to any one of the above aspects or an embodiment thereof.

BRIEF DESCRIPTION OF THE DRAWINGS

FIGURE 1 illustrates a network scenario in accordance with at least some embodiments;

FIGURE 2 illustrates a machine- learning architecture for wireless nodes;

FIGURE 3 illustrates a first method in accordance with at least some embodiments;

FIGURE 4 illustrates a second method in accordance with at least some embodiments;

FIGURE 5 a and 5b are signalling diagrams in accordance with at least some embodiments; and

FIGURE 6 illustrates an example apparatus capable of supporting at least some embodiments.

EMBODIMENTS

FIGURE 1 illustrates a simplified example system in accordance with at least some embodiments. A user equipment (UE) 10 communicates wirelessly with a wireless radio or access network node, hereafter referred to as AN, 20, such as a NodeB, an evolved NodeB (eNB), a Next Generation (NG) NodeB (gNB), a base station, an access point, or other suitable wireless/radio access network device or system.

The UE 10 may be attached to a cell of the AN, 20, 30 for wireless communications. For example, at time instant tl, the AN 20 may be a serving AN or serving cell for the UE 10. The air interface between UE and AN may be configured in accordance with a Radio Access Technology, RAT, which both the UE 10 and AN 20, 30 are configured to support. Examples of cellular RATs include Long Term Evolution, LTE, New Radio, NR, which is also known as fifth generation, 5G, and MulteFire. On the other hand, examples of non-cellular RATs include Wireless Local Area Network, WLAN, and Worldwide Interoperability for Microwave Access, WiMAX. Principles of the present disclosure are not limited to a specific RAT though. For example, in the context of LTE, AN 20, 30 may be a nodeB or evolved Node B (eNB), while in the context of NR, AN 20, 30 may be a gNB, as in some of the example embodiments below.

The AN 20, 30 may be connected, directly or via at least one intermediate node, with a core network (not shown), such as a Next Generation core network, Evolved Packet Core (EPC), or other network management element.

The radio access network and the core network may comprise a set of network functions. A network function may refer to an operational and/or physical entity. The network function may be a specific network node or element, or a specific function or set of functions carried out by one or more entities, such as virtual network elements. Examples of such network functions include a (radio) access or resource control or management function, session management or control function, interworking, data management or storage function, authentication function or a combination of one or more of these functions.

For example, a 5G core network comprises Access and Mobility Management Function (AMF) which may be configured to terminate radio access network (RAN) control plane (N2) interface and perform registration management, connection management, reachability management, mobility management, access authentication, access authorization, Security Anchor Functionality (SEAF), Security Context Management (SCM), and support for interface for non-3GPP access. The AMF is in charge for managing handovers between gNBs.

The core network may be, in turn, coupled with another network (not shown), via which connectivity to further networks may be obtained, for example via a worldwide interconnection network. The AN may be connected with at least one other AN as well via an inter-base station interface, particularly for supporting mobility of the UE 10, e.g. by 3 GPP X2 or similar NG interface. The UE 10 may be referred to as a user device or a wireless terminal in general. Hence, without limiting to Third Generation Partnership Project (3GPP) User Equipment, the term user equipment or UE is to be understood broadly to cover various mobile/wireless terminal devices, mobile stations and user devices for user communication and/or machine to machine type communication. The UE 10 may be or be comprised by, for example, a smartphone, a cellular phone, a Machine-to -Machine, M2M, node, machine-type communications node, an Internet of Things, IoT, node, a car telemetry unit, a laptop computer, a tablet computer or, indeed, another kind of suitable user device or mobile station, i.e., a terminal.

The example system of FIGURE 1 may apply a set of beams to provide cell coverage. For example, if the system is a 5G mmwave system, a grid of beams is applied and it is known at each instance in time which beam is serving a given user. The UE 10 may be configured to report channel measurements, such as measurements on a number of beams, via channel state information (CSI). Based on the reported channel measurements, the AN 20 can perform cell selection and/or beam management procedure. For example, the AN 20 may determine which is the best beam that should serve the UE 10, and perform a beam management procedure signaling the UE 10 to receive from the best beam.

In some embodiments, the AN 20, or a further network entity connected to the AN, may be configured to apply a machine learning (ML) model and store history data of which beam best serves the UE 10 as the UE continues on its trajectory. For example, when the UE’s signal to the serving AN 20 is suddenly blocked or weakened when moving from time instant tl to t2 and a handover request is generated, the ML model can be trained to predict when these blockage events happen. The model may be trained to predict to which new cell ID the UE needs to be handed over to, e.g. based on past time series of beam indices. Handover refers herein generally to change of serving wireless access network node, in cellular systems change of serving cell.

After the ML model has been sufficiently trained, it may be deployed to make predictions e.g. on need for handover. Such deployment stage may also be referred to as online inference stage or phase. A prediction algorithm can be trained and deployed based on collected measurements in a network infrastructure node, such as AN 20, a combination of multiple network infrastructure nodes, or another node in the network. If the ML model predicts with high probability that a blockage event is going to occur and a handover to a new AN will take place, the system can proactively issue a handover command to the UE 10 prior to the actual blockage event happening. This enables to achieve better performance as the sudden loss in signal quality can be avoided and a high quality of service (e.g. high data rate and/or low packet loss rate) can be maintained for the UE. It is to be noted that the system may be configured to further train the model also when it is being deployed, to react to changes in the environment.

Certain ML-based applications may also be run in the UE 10. This may allow the UE to take more/better decisions autonomously and provide more accurate or timely requests to the RAN, such as the serving AN 20. This also means that control and feedback signalling between the RAN and the UE would need to be adapted to the envisioned operating conditions of ML-based prediction algorithms running in the RAN and/or UE.

A simplified example for a possible functional architecture for the scenario when both the AN 20 and the UE 10 have ML-based functionalities is illustrated in FIGURE 2. The training and inference functionalities are separated to illustrate that these can be generally decoupled and active either in parallel or at different time instances. Training generally implies learning of optimized parameters of the ML model, such as weights of neural networks, based on training data available, with the aim to create a model that generalizes well to unknown data set. Inference implies that data-driven predictions or decisions are taking place using the trained ML model. Examples of further learning models include un-supervised learning or self-supervised learning algorithms.

The depicted control-feedback message loop indicates a generic set of information which may be exchanged between the ML-enabled RAN (gNB) and the ML- capable UE: configuration of the ML-based fimctionality(ies), e.g. selection of the ML model, and reporting of inference results.

Different ML-based architecture options may be applied for radio resource management (RRM). In modular ML architecture, one or more ML instances or modules may be attached to RRM function of serving AN 20 and/or served UEs 10 and may perform inference of configured key-performance indicators (KPI(s)). In AN ML-based assistance, the AN 20 may be capable of generating inference based on feedback from the UE 10 and other data collected at the network. The inference results may be used to take RRM actions and/or to assist the RRM actions at the served UEs. In UE ML-based assistance, the UE may be capable of generating inference for assisting the serving AN 20 to take RRM actions or improve ML training, e.g. during exploration phase. Employed ML algorithms may need to be continuously (or regularly) retrained or fine-tuned and validated while the network is running, and the radio conditions are changing.

The UE 10 may thus comprise the functionalities needed to provide ML-based assistance to the RAN, such as the AN 20. These functionalities may comprise a prediction algorithm configured to generate inference information. The inference information may be, for example, on one or more of: prediction (or identification) of certain events, such as handovers, signal thresholds (e.g. reference signal received power (RSRP) values, quality of service (QoS) variations, channel state information (CSI), and mobility state change. The inference information then needs to be reported back to the serving AN 20 to be used as input in the RRM algorithms and RRM actions. A reporting mechanism is needed for providing ML-based inference results. The reporting mechanism should also be configurable, in some embodiments by the AN.

There are now provided further improvements facilitating to better interpret received inference results for radio access control, such as proactive mobility management. The improvements include a procedure for providing, from inference-generating wireless device, referred herein as the first wireless device, to inference-consuming wireless device (second wireless device), information that may have affected the processing resulting the inference.

FIGURE 3 is a flow graph of a first method in accordance with at least some embodiments. The illustrated first method may be performed by the first wireless device, such as a user equipment, e.g. the UE 10, or alternatively a network node, e.g. the AN 20, or by a control device configured to control the functioning thereof, possibly when installed therein. It is to be noted that an action, such as transmitting, in a given block may refer to controlling or causing such action in another apparatus or unit.

The method comprises determining 300 at least one parameter of a prediction algorithm. This may refer to the training stage of a ML model of the prediction algorithm or to determining the parameter(s) to be applied in the ML model, e.g. based on a received configuration control message. An input is applied 310 to the prediction algorithm with the at least one determined parameter to obtain a first inference result for radio access control function. The radio access control function refers generally to a function in a network node or a terminal for controlling radio access related operation, which may include mobility management control, for example.

Block 320 comprises transmitting to a second wireless device inference conditions information associated with the first inference result and characterizing the conditions under which the first inference result was obtained.

FIGURE 4 is a flow graph of a second method in accordance with at least some embodiments. The illustrated method may be performed by the second wireless device, such as the AN 20 or the UE 10, or by a control device configured to control the functioning thereof, possibly when installed therein. The second wireless device may be configured to communicate with the first wireless device performing the method of FIGURE 3.

The method comprises receiving 400 from the first wireless device inference conditions information associated with a first inference result and characterizing conditions under which the first inference result was obtained by a prediction algorithm in the first wireless device.

A second inference result is generated 410 on the basis of the first inference result and the inference conditions information for controlling a radio access function on the basis of the second inference result. The radio access function may then be controlled on the basis of the second inference result, by the apparatus performing the method or another apparatus receiving the information second inference result after block 410. For example, the second inference result may comprise a prediction that handover or beam change is needed, on the basis of which handover or beam change may be initiated proactively before the signal gets weak.

Inference refers generally to an output of a prediction algorithm, such as a neural network based algorithm, which may be configured to deploy an ML model. The inference may be a direct output from the prediction algorithm or the prediction algorithm output after one or more subsequent processing steps. The inference may be predictive and comprise a prediction and/or causal processing result, which may comprise analyzing or interpreting a prediction result by a prediction algorithm. For example, the (first) inference result from the UE 10 may be input to a prediction algorithm in the AN 20, producing a further (second) inference based on the first inference and other inputs. In some embodiments, the prediction algorithm is configured to identify one or more types of events, such as an event due to mobility of the UE 10. The term prediction algorithm is thus to be understood broadly and may comprise an algorithm employing an ML model for event identification or decision making (even without prediction or forecasting), and the inference or another output of the prediction algorithm may comprises identification of an event.

The inference conditions information may also be referred to as context information and refers generally to information that may have affected the processing resulting the first inference result by the prediction algorithm, and may be comprise cause values. The inference conditions information may be configured to provide sufficiently accurate temporal and spatial description of the affecting conditions. The inference conditions information includes at least one parameter. The inference conditions information may be a combination of several existing and/or new cause indicators.

In some embodiments, the inference conditions information comprises radio access parameter information not known a priori by or otherwise being reported to the second wireless device. Thus, the inference conditions information is not a CSI feedback report or about radio channel measurement triggers, events or conditions known by the second wireless device. The inference conditions information may classify or characterize at least one of: side information available to the first wireless device, the prediction algorithm, and an input to the prediction algorithm. The side information refers generally to information not input of or about the prediction algorithm (and thus not associated with the algorithm), but may affect the inference result. The inference conditions information may be indicative of at least one type of input to the prediction algorithm.

In an embodiment, inference conditions information is obtained from the prediction algorithm itself. Thus, the ML model may be trained to identify most relevant cause(s) to obtaining the first inference result. The prediction algorithm may be configured to provide such cause output information in addition to the first inference result.

The inference conditions information may indicate one or more of the following aspects associated with the generation of the first inference result: accuracy of validation of the first inference result, inference operating mode, all inference values obtained from the prediction algorithm included or not included, and importance of one or more included conditions.

For example, there may be (at least) two inference operating modes: exploration mode and normal inference mode. Indication of the operating mode may be relevant e.g. in case the exploration mode is configured to provide a specific set of output KPIs. In reinforcement learning mode some of the inference results may be selected randomly due to required exploration phase. The exploration operating mode may refer to an operating mode where the UE has been configured to react (e.g. measure, estimate and report) certain metrics based on specially designed, or tagged, DL signals. The RRM action based on this exploration operating mode are not affecting the normal operating mode of the UE, and reports comprising the inference conditions information and provided to the network may assist the network to determine how and why UE’s ML model works and performs.

An inference result may be selected from two or more output values (which may be from different ML models or entities) by a decision algorithm. The inference report may include only the selected output value or all inferred output values within a predefined time window. The time window and/or the selection may be indicated in the inference conditions information.

In case of reinforcement learning, an ML-based algorithm may be configured to evaluate and apply a cost function for actions or outcome of or resulting from the inference result. The algorithm, such as the prediction algorithm discussed above, may then adapt its parameters on the basis of the estimated cost.

In a still further example embodiment, the inference conditions information comprises ML model or architecture information. The ML model information may comprise one or more of an identifier of the applied ML model, information on weights of the ML model, indication of activation functions, and other parameters of the ML model. Lor example, the ML model information may comprise or indicate self-test or self-validation accuracy of the ML model.

It is to be noted that if the inference report to the second wireless device comprises all output values (features) of the prediction algorithm, the accuracy (of the applied ML model by the prediction algorithm) can also be validated by the second wireless device on the basis of the outputs. In such cases, the accuracy it does not need to be reported as part of the inference conditions information. On the basis of the inference conditions information provided by the methods, the second wireless device is better able to ‘interpret’ the received inference result correctly. The second wireless device may on the basis of the received inference conditions information determine if the received inference result is to be ’trusted’ and/or if the inference result requires additional processing (filtering, etc.) or additional actions, such as activation of exploration is to be triggered. A configurable and hierarchical ML-specific context information signaling mechanism may thus be provided between RAN entities or nodes, which may be adapted to various RAN level prediction algorithm implementations.

In general, causality of certain input-output pair may be very difficult to establish for ML-based algorithms, and may be possible only in very limited number of cases, especially in the context of radio access control related algorithms. Sufficient amount of inference conditions information may enable to establish such causality, along with dedicated algorithms which aim to provide explainability and interpretability. Devices employing ML-based entities, such as gNBs, UEs, and/or CN entities in a 5G system, can leam over time the essential context conditions for particular inference results. This information can help design algorithms dedicated to provide explainability and interpretability of the inference results, which in turn can be used to improve radio resource management actions and decisions.

Furthermore, reporting of certain inference conditions information for several consecutive inference results which are later determined to be erroneous, may provide indication to the second wireless device on if (and when) exploration or training needs to be triggered again. This may be because inference data is very different from the trained data, e.g. in the case of exploration operation mode. The second wireless device may thus based on block 410 initiate exploration or training operation in the second wireless device and/or the first wireless device.

Some further embodiments are illustrated below, comprising references also to the UE (10) and AN (20) representing the first wireless device and the second wireless device, respectively. However, it is to be appreciated that at least some of the embodiments may be applied in the case where an access network node implements the method of Figure 3 and a user device implements the method of Figure 4. Further, it is to be noted that the presently disclosed inference conditions information may be provided from a receiving device to other devices. For example, the received inference conditions information may be transmitted from the AN 20 to another access network node or a core network element, or from one UE to another.

The inference condition information may be determined by the UE 10 based on predefined one or more types of inference condition information, or based on determining a new type of inference condition information. Available inference condition information may be initially determined or detected by the UE 10 in an initial configuration phase or when ML-related capabilities are exposed or advertised. In such exposed or advertised information, each inference condition information type may be assigned type identification.

The prediction algorithm may be able to even identify a new root cause, which have not been predefined in the training stage. Such newly identified root causes may be added as new inference conditions information types and included in the inference conditions information transmitted 320, 400 to the AN 20 in block 320. The AN 20 may be configured to detect and process also new inference conditions information types.

The hierarchical configuration and/or activation/deactivation of the inference conditions information signalling may be arranged by dedicated or group control signalling from the inference consumer entity to the entity generating the inference, i.e. from the AN 20 to the UE 10. Since there may be a high number of available inference conditions information types and/or reporting types available, the UE 10 and/or the AN 20 may be configured to dynamically limit the number of types and thus amount of inference conditions information being reported to reduce signalling burden.

The inference conditions information may be included in an inference report comprising the first inference result. Alternatively, the inference conditions information may be transmitted independently of and separately from the first inference result in a dedicated inference conditions information signalling or report. The inference conditions information may be added as new information element(s) in an existing control message, such as one or more 5G RRM (or radio resource control (RRC)) messages, or a new message type may be specified for the inference conditions information (and/or the first inference result). For example, the inference conditions information may be included in one or more of an RRC complete message, an RRC disconnect message, and a dedicated RRC condition or context report message. Transmission of the inference conditions information from the UE 10 may be triggered (by the AN 20, or even another network node) and/or periodic. In some embodiments, the AN 20 is configured to transmit to the UE 10 a control message comprising control parameter(s) for inference conditions information reporting. The UE 10 adapts inference conditions information reporting for block 320 on the basis of the received control parameters.

Such control message may comprise types of inference conditions information to be reported for the AN 20. The UE 10 obtains or determines the inference conditions information in accordance with the received control message. Each inference condition information type may be associated with a list of supported (or not-supported) inference conditions information reporting types or configurations. For example, areas with good coverage and low network load, an optimum HO point for a certain UE might be of lower relevance for a gNB.

There may be a set of inference conditions reporting types, among which one or more reporting types are selected for block 320. The control message from the AN 20 may comprise an indication of an inference conditions reporting type, and the inference conditions information is included in a report in accordance with the received inference reporting type. In case a new inference conditions information type is identified, a new inference reporting type may also be established and added in the list of available inference types/reports in a predefined manner so that it can be used in future inference reports.

An (uplink) control channel message from the UE 10 to the AN 20 may be used to indicate if inference conditions information will be provided for certain (uplink) inference conditions information report. The control channel message may further comprise information of the inference conditions information types and/or inference conditions information reporting types to be provided. Thus, the UE 10 may proactively inform the AN 20 about inference conditions information to be reported, or being reported in case such characterizing information is transmitted at the same time as the inference conditions information.

The AN 20 may apply the received inference conditions information in various ways. It may receive inference results and associated inference conditions information from one or more further wireless devices and generate the second inference result in block 410 on the basis of inference results and associated inference conditions information from a plurality of wireless devices. Some further example embodiments and use cases are illustrated below.

The AN 20 may evaluate accuracy or reliability of the first inference result on the basis of the inference conditions information and further information available to the AN 20. As an example, prediction reports associated with certain conditions, received from certain UE may consistently cause generating a ping-pong type of HO. This is detectable by the network, thus the UE ML assistance can be either disabled or just flagged as unreliable. In case the above condition only holds for that UE when in specific area, e.g. as in Figure 1, then the reliability of the inference conditions information can be explored. In case the above condition hold for several UEs, all located in the same area, then the network can decide to disable ML for all UEs in the area or treat their predictions/condition as unreliable; this can then be followed by other algorithms aimed to remedy network performance.

In some embodiments, the radio access control function comprises a handover or beam management function and the inference results comprise a radio cell handover prediction or a beam selection prediction, which may indicate a predicted best beam. For example, the inference condition information may be transmitted as a part of a handover request.

In an example embodiment, movement of the UE 10 is detected. The UE may be configured to include in the inference conditions information an indication of its detected movement, and possible further characterizing information on the movement, such as speed. This information may be applied as an input by the AP 20, e.g. an ML-based proactive mobility management module thereof, to predict need for mobility management actions.

The inference conditions information may comprise indication of radio access related parameters applied for generating the first inference result and/or affecting the generation of the first inference result. In some embodiments, the inference conditions information indicates if the prediction algorithm was input by one or more of: signal to noise (SNR) inference ratio to handover source cell and handover target cell, interference level from other cells, movement velocity considered or not considered, exploration active or not active, best beam identifier, antenna port signal to noise ratio, inter cell cross talk, sequence of applied antenna panels. Above-illustrated parameters and possible further parameters may be applied and indicated in the inference conditions information in various combinations, depending on the desired application. For example, the first device and the second device may be UEs, in which case the inference conditions information may indicate one or more of signal-to- interference-plus-noise ratio (SINR) of a sidelink (device-to-device (D2D) link), inference level on the sidelink, and signal strength characteristics on the sidelink. In another example, the first device and the second device may be wireless network nodes, such as gNBs. Thus, the inference conditions information may indicate one or more of DL and/or UL traffic load, traffic type, and traffic profile.

The UE 10 may be configured to apply the prediction algorithm to predict its radio active periods. The UE 10 may be configured to use side information not available in the radio access network, e.g. in the AN 20 (and not sensible to report to RAN as such). For example, UE’s internal sensors, processes and applications may provide information that correlates with UE’s probability to initiate UL traffic. Therefore, sending activity prediction as the first inference result associated with the inference conditions information such as validation accuracy of the inference, the AN 20 (or another node in the network) may decide if to trust the inference and trigger actions to prepare the upcoming UL transmission.

In some embodiments, at least some of the presently disclosed features are applied for 5G NR systems, some further such embodiments being illustrated below. In an example, with reference to FIGURES 1, 5a, and 5b, it is assumed that AN 20 is a gNB consumer of inference reports generated by an ML-based prediction algorithm in the UE 10. The AN 30 may be or provide a secondary cell (SC). This is a typical heterogeneous mobile radio scenario (or small part of it), where the SC is HO candidate for the UE connected to the gNB.

For example, in FIGURE 1, due to objects 40, 42, e.g. buildings, a beam 32 of the SC/AN 30 may be shadowed for the UE 10 moving upwards in the direction of the arrow at time instant tl, then at t2 the UE receives the beam 32 with strong power, and shortly afterwards at t3 the beam 32 will be weak once again.

As illustrated in the signaling example of FIGURE 5a, the UE may receive CSI references signals 500, 502 from the gNB and the SC. The UE 10 may be configured to determine 504 and transmit an inference report 506, which may comprise, in addition to the inference result for requesting a HO or not, an ML-context (as the inference conditions information), which led to its inference. As already illustrated, such ML-context could be indication of input measurements or input events associated with the ML inference, information about an ML model (applied by the prediction algorithm) that was used for the inference, and/or side information not associated with the ML model.

The UE 10 may be configured, by applying the inference at 504, to detect a need to make a HO to the SC at certain predicted time instance in the future if it estimates that it will receive the SC with much stronger power than the gNB. Then the UE may report at 506 to the gNB the inference result as a request to perform a HO to the SC at a future time instance X (e.g. a frame number).

In addition, the UE sends ML-context (as the inference conditions information) in the inference report 506, or may have at least partially already provided it to the gNB already earlier. The ML-context may be in a form of an ML-related feature importance score list applied for the current ML model, for example. For example, if the UE prediction algorithm is using as input not only estimated signal levels but also the CSI and detected movement velocity, then the feature importance score list could indicate that the particular HO (and occurrence time X) was inferred in the context of the CSI and velocity rather than only the RX signal levels from the SC and gNB. An additional indication on the condition with strong RX-power of the SC (at t2) may also be included in the inference conditions information report.

The gNB may generate 508 the second inference result on the basis of the received inference and ML-context from the UE 10, as well as further inputs, such as inference and inference conditions information from other UEs passing the objects 40, 42, and/or HO inference based on channel measurements by the gNB. This may lead to the gNB to learn not to perform HO for UEs to the SC passing the objects 40 and 42, as the SC will be out of reach again very soon.

FIGURE 5 a also further illustrates another case where a handover is required, in which case the AN 20 may transmit a HO decision or control message 510, in response to which the UE 10 may initiate handover 512 to the SC, for example. FIGURE 5b illustrates another example embodiment, where the gNB transmits an inference result reporting configuration 550 to the UE for controlling a report comprising the first inference result. Such report may be the inference report like in FIGURE 5 a, or a specific inference result report 556 transmitted separately from the inference conditions information.

The gNB may also transmit an inference conditions reporting configuration 552 for controlling the transmission of the inference conditions information, which may be included in an inference conditions report 558. This configuration 552 may comprise some or all of the information illustrated above in connection with the control message, such as the inference information type(s) and/or inference information reporting type(s) to be applied. It is to be further noted that the configurations 550 and 552 may be alternatively transmitted together in a single control message. Block 554 illustrates generation of the inference result and/or definition of the inference conditions information. In block 560 the gNB may process received one or more inference result reports 556 and associated inference conditions information report(s) 558 and decide to perform an RRM action, which may be communicated to the UE by an RRM action message 562. Thus, an inference conditions result report and the inference conditions information thereof may be applied for processing a plurality of inference result reports.

In case an exploration operating mode is activated in a ML capable UE, inferred exploration-HOs might not need to be executed in practice. The inference results and associated inference conditions information may be used by the gNB to configure or teach other RRM algorithms, such as load balancing.

A more reliable combined inference is thus possible by applying inference conditions information from a plurality of wireless devices. As a further example, the gNB, the UEs, and potentially even the SC may have their own ML instances, i.e., gNB instance #1 and UE instance #2 for generating the inference about a potential handover to the SC.

The gNB and the UE may have substantially different information available affecting and input to the inference generation, e.g. the gNB might have access to a full and detailed building vector data map (BVDM), the UE location (if reported), the scheduler decisions over latest transmission time intervals (TTIs), the according cell specific load conditions, the configuration of CSI RSs, the TX power, the beam pattern design, etc. Contrarily, the UE may use the CSI RSs for constant channel estimation to its own cell, RSRP measurements from time to time to other candidate cells, interference measurements, e.g., based on interference measurement resources, etc. The two different ML instances could come up with the different inference results regarding a decision at the same time instance. In reality, we can expect a certain variance for the inference results. By providing the inference conditions information the underlying affecting information may be taken into account in the decision making and/or inference generation at the inference information receiving entity.

It is to be noted that handover was above used as an example for illustration particularly due to its simplicity. However, besides the already mentioned ones, there are several other available applications for the inference conditions information, such as beam management for mmWave frequency range, dual connectivity carrier aggregation (DCCA), pre-coding matrix indicator (PMI), channel quality indicator (CQI) reporting, network operation and maintenance (O&M), and/or network slicing management, for which the inference conditions information may be provided in a similar manner.

The inference conditions information may be transmitted in a structured information format, in an example embodiment in an approach and format similar to measurement objects approach currently specified in 3GPP LTE and 5G. Some examples on structure and information elements for inference conditions information are illustrated in Tables 1 to 3. The inference conditions information is herein referred to as ML-context (mlcontext). ML-context reporting functionalities may be linked to the corresponding ML- based inference capabilities information.

Table 1.

Table 2.

Table 3.

For example, during UE capability query-response phase, the UE 10 may be configured to expose its ML-based capability(ies) and/or functionality(ies). Some or all of the ML-context report types illustrated in Tables 1 and 3 may be identified as inference conditions information for HO prediction and some or all of the types illustrated in Table 2 may be selected as inference conditions information for beam selection prediction. The Inference Type indicates the type of the inference result or output. For example, for a HO prediction the type may be simply true or false, or yes/no.

The ML-context reports may be embedded in associated ML-based HO or beam selection inference reports, or transmitted separately. For example, the information element mlcontextReportType#3 may indicate importance scores of ML features for HO prediction, e.g. as extracted via a random forest regression model. Such context report may be included in an associated HO selection inference report. The mlcontextReportType#6 may provide context for gNB beam selection predictions by sequence of the used UE antenna panels (e.g. by listing their IDs). Such context report may be sent separately from ML-based beam selection prediction reports.

The gNB may transmit a configuration ( mlcontextConfig ) to the UE for the inference conditions information reporting by indicating one or more ML-context objects (; mlcontextObject ) to be configured and activated. For example mlcontextObject#! may comprise at least some of the below three reports #1 to #3 (and thus indicate such report configuration(s)) :

Context Report configuration ID: 1 Context Report Type: mlcontextReportType#3 (refers to context ID for HO inference)

Context Reporting mode: triggered (may be triggered by HO execution) Context Reporting format: 500ms - beam-wise (long term estimate for each detected gNB radio beam. 500ms is an example of the periodicity of reporting the context information. “Beam-wise” may indicate that the reports should include context information derived for each of the detected radio beams (i.e. based on beam measurements).) mlcontextReport #2:

Context Report configuration ID: 2

Context Report Type: mlcontextReportType#3 (context ID for HO inference)

Context Reporting mode: periodic- 100ms (periodic report in this example with 100ms interval)

Context Reporting format: 100ms - cell-wise (long term estimate for each detected cell. Cell-wise may indicate context to be reported per detected cells (i.e. based on cell measurements).) mlcontextReport #3:

Context Report configuration ID: 3

Context Report Type: mlcontextReportType#6 (context ID for Beam Selection inference)

Context Reporting format: 10ms - cell-wise (short term estimate for each detected cell)

As further additions to Table 1, the inference could be associated with certain ML model information, such as one or more types illustrated in Table 3. For example, with predictive HO, by reporting validation accuracy ( wilContextReportType#8 in Table 2) and SINR ( mlContectReportType #1 in Table 1) along with the HO inference report, the network would be able to evaluate quality of goodness of the ML model and signal conditions. For instance, network may have a separate algorithm (possibly ML-based) to verify suitability of the predicted HO with respect to the reported SINR values, and decide when to trust, or not, the HO inference reports.

Presently disclosed features maybe particularly beneficial for 5G Ultra Reliable Low Latency Communications (URLLC), for which it is particularly important to ensure that the UE does not have its transmission interrupted by a blocking event. However, it will be appreciated that this is just one example and at least some of the above illustrated features may be applied in various cellular or non-cellular wireless systems.

An electronic device comprising electronic circuitries may be an apparatus for realizing at least some embodiments of the present invention. The apparatus may be or may be comprised in a computer, a laptop, a tablet computer, a cellular phone, a machine to machine (M2M) device (e.g. an IoT sensor device), a wearable device, a base station, access point device, a radio access control device, a network function element or node, or any other apparatus provided with radio access control capability. In another embodiment, the apparatus carrying out the above-described functionalities is comprised in such a device, e.g. the apparatus may comprise a circuitry, such as a chip, a chipset, a microcontroller, or a combination of such circuitries in any one of the above-described devices.

FIGURE 6 illustrates an example apparatus capable of supporting at least some embodiments of the present invention. Illustrated is a device 600, which may comprise a communications device arranged to operate as the AN 20, or the UE 10, for example. The device may include one or more controllers configured to carry out operations in accordance with at least some of the embodiments illustrated above, such as some or all of the features illustrated above in connection with FIGURES 1 to 5b. The device may be configured to operate as the apparatus configured to perform the method of FIGURE 3 and/or 4, for example.

Comprised in the device 600 is a processor 602, which may comprise, for example, a single- or multi-core processor wherein a single-core processor comprises one processing core and a multi-core processor comprises more than one processing core. The processor 602 may comprise more than one processor. The processor may comprise at least one application- specific integrated circuit, ASIC. The processor may comprise at least one field-programmable gate array, FPGA. The processor may be means for performing method steps in the device. The processor may be configured, at least in part by computer instructions, to perform actions.

A processor may comprise circuitry, or be constituted as circuitry or circuitries, the circuitry or circuitries being configured to perform phases of methods in accordance with embodiments described herein. As used in this application, the term “circuitry” may refer to one or more or all of the following: (a) hardware-only circuit implementations, such as implementations in only analog and/or digital circuitry, and (b) combinations of hardware circuits and software, such as, as applicable: (i) a combination of analog and/or digital hardware circuit(s) with software/firmware and (ii) any portions of hardware processor(s) with software (including digital signal processor(s)), software, and memory(ies) that work together to cause an apparatus, such as a mobile phone or server, to perform various functions) and (c) hardware circuit(s) and or processor(s), such as a microprocessor(s) or a portion of a microprocessor(s), that requires software (e.g., firmware) for operation, but the software may not be present when it is not needed for operation.

This definition of circuitry applies to all uses of this term in this application, including in any claims. As a further example, as used in this application, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and/or firmware. The term circuitry also covers, for example and if applicable to the particular claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in server, a cellular network device, or other computing or network device.

The device 600 may comprise memory 604. The memory may comprise random- access memory and/or permanent memory. The memory may comprise at least one RAM chip. The memory may comprise solid-state, magnetic, optical and/or holographic memory, for example. The memory may be at least in part accessible to the processor 602. The memory may be at least in part comprised in the processor 602. The memory 604 may be means for storing information. The memory may comprise computer instructions that the processor is configured to execute. When computer instructions configured to cause the processor to perform certain actions are stored in the memory, and the device in overall is configured to run under the direction of the processor using computer instructions from the memory, the processor and/or its at least one processing core may be considered to be configured to perform said certain actions. The memory may be at least in part comprised in the processor. The memory may be at least in part external to the device 600 but accessible to the device.

For example, computer program code and control parameters for performing the prediction algorithm and affecting controlling operations illustrated in connection with Figures 3 and/or 4 may be stored in one or more portions of the memory and used to control operation of the apparatus. Further, the memory may be used as a temporary data storage, e.g. for prediction algorithm input data and inference results.

The device 600 may comprise a transmitter 606. The device may comprise a receiver 608. The transmitter and the receiver may be configured to transmit and receive, respectively, information in accordance with at least one wired or wireless, cellular or non- cellular standard. The transmitter may comprise more than one transmitter. The receiver may comprise more than one receiver. The transmitter and/or receiver may be configured to operate in accordance with global system for mobile communication, GSM, wideband code division multiple access, WCDMA, long term evolution, LTE, 5G or other cellular communications systems, WLAN, and/or Ethernet standards, for example. The device 600 may comprise a near- field communication, NFC, transceiver 610. The NFC transceiver may support at least one NFC technology, such as NFC, Bluetooth, Wibree or similar technologies.

The device 600 may comprise user interface, UI, 612. The UI may comprise at least one of a display, a keyboard, a touchscreen, a vibrator arranged to signal to a user by causing the device to vibrate, a speaker and a microphone. A user may be able to operate the device via the UI, for example to accept incoming telephone calls, to originate telephone calls or video calls, to browse the Internet, to manage digital files stored in the memory 604 or on a cloud accessible via the transmitter 606 and the receiver 608, or via the NFC transceiver 610, and/or to configured the operation and parameters of the device.

The device 600 may comprise or be arranged to accept a user identity module or other type of memory module 614. The user identity module may comprise, for example, a subscriber identity module, SIM, and/or a personal identification IC card installable in the device 600. The user identity module 614 may comprise information identifying a subscription of a user of device 600. The user identity module 614 may comprise cryptographic information usable to verify the identity of a user of device 600 and/or to facilitate encryption and decryption of communication effected via the device 600.

The processor 602 may be furnished with a transmitter arranged to output information from the processor, via electrical leads internal to the device 600, to other devices comprised in the device. Such a transmitter may comprise a serial bus transmitter arranged to, for example, output information via at least one electrical lead to memory 604 for storage therein. Alternatively to a serial bus, the transmitter may comprise a parallel bus transmitter. Likewise the processor may comprise a receiver arranged to receive information in the processor, via electrical leads internal to the device 600, from other devices comprised in the device 600. Such a receiver may comprise a serial bus receiver arranged to, for example, receive information via at least one electrical lead from the receiver 608 for processing in the processor. Alternatively to a serial bus, the receiver may comprise a parallel bus receiver.

The device 600 may comprise further devices not illustrated in FIGURE 6. For example, the device may comprise at least one digital camera. Some devices may comprise a back-facing camera and a front-facing camera. The device may comprise a fingerprint sensor arranged to authenticate, at least in part, a user of the device. In some embodiments, the device lacks at least one device described above. For example, some devices may lack the NFC transceiver 610 and/or the user identity module 614.

The processor 602, the memory 604, the transmitter 606, the receiver 608, the NFC transceiver 610, the UI 612 and/or the user identity module 614 may be interconnected by electrical leads internal to the device 600 in a multitude of different ways. For example, each of the aforementioned devices may be separately connected to a master bus internal to the device, to allow for the devices to exchange information. However, as the skilled person will appreciate, this is only one example and depending on the embodiment various ways of interconnecting at least two of the aforementioned devices may be selected without departing from the scope of the present invention.

It is to be understood that the embodiments of the invention disclosed are not limited to the particular structures, process steps, or materials disclosed herein, but are extended to equivalents thereof as would be recognized by those ordinarily skilled in the relevant arts. It should also be understood that terminology employed herein is used for the purpose of describing particular embodiments only and is not intended to be limiting.

Reference throughout this specification to one embodiment or an embodiment means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, appearances of the phrases “in one embodiment” or “in an embodiment” in various places throughout this specification are not necessarily all referring to the same embodiment. Where reference is made to a numerical value using a term such as, for example, about or substantially, the exact numerical value is also disclosed.

As used herein, a plurality of items, structural elements, compositional elements, and/or materials may be presented in a common list for convenience. However, these lists should be construed as though each member of the list is individually identified as a separate and unique member. Thus, no individual member of such list should be construed as a de facto equivalent of any other member of the same list solely based on their presentation in a common group without indications to the contrary. In addition, various embodiments and example of the present invention may be referred to herein along with alternatives for the various components thereof. It is understood that such embodiments, examples, and alternatives are not to be construed as de facto equivalents of one another, but are to be considered as separate and autonomous representations of the present invention.

In an exemplary embodiment, an apparatus, such as, for example, a user equipment or terminal or a network node, may comprise means for carrying out the embodiments described above and any combination thereof.

In an exemplary embodiment, a computer program may be configured to cause a method in accordance with the embodiments described above and any combination thereof. In an exemplary embodiment, a computer program product, embodied on a non-transitory computer readable medium, may be configured to control a processor to perform a process comprising the embodiments described above and any combination thereof.

In an exemplary embodiment, an apparatus, such as, for example, a terminal or a network node, may comprise at least one processor, and at least one memory including computer program code, wherein the at least one memory and the computer program code are configured to, with the at least one processor, cause the apparatus at least to perform the embodiments described above and any combination thereof.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the preceding description, numerous specific details are provided to provide a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention can be practiced without one or more of the specific details, or with other methods, components, materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the invention.

While the forgoing examples are illustrative of the principles of the present invention in one or more particular applications, it will be apparent to those of ordinary skill in the art that numerous modifications in form, usage and details of implementation can be made without the exercise of inventive faculty, and without departing from the principles and concepts of the invention. Accordingly, it is not intended that the invention be limited, except as by the claims set forth below. The verbs “to comprise” and “to include” are used in this document as open limitations that neither exclude nor require the existence of also un-recited features. The features recited in depending claims are mutually freely combinable unless otherwise explicitly stated. Furthermore, it is to be understood that the use of "a" or "an", that is, a singular form, throughout this document does not exclude a plurality.

Claims

CLAIMS:

1. An apparatus for a first wireless device, comprising means for performing:

- determining at least one parameter of a prediction algorithm;

- applying an input to the prediction algorithm with the at least one determined parameter to obtain a first inference result for radio access control function; and

- transmitting to a second wireless device inference conditions information associated with the first inference result and characterizing the conditions under which the first inference result was obtained.

2. The apparatus of claim 1, wherein the means are further configured for: receiving a control message from the second wireless device, indicative of one or more types of inference conditions information to be provided for the second wireless device, and obtaining the inference conditions information in the first wireless device in accordance with the received control message.

3. The apparatus of claim 1, wherein the means are configured for determining the inference condition information based on predefined one or more types of inference condition information, or based on determining a new type of inference condition information.

4. The apparatus of any preceding claim, wherein the means are further configured for receiving an inference result reporting configuration from the second device and generating a report comprising at least the first inference result in accordance with the received inference result reporting configuration.

5. An apparatus for a second wireless device, comprising means for performing:

- receiving from a first wireless device inference conditions information associated with a first inference result and characterizing conditions under which the first inference result was obtained by a prediction algorithm in the first wireless device, and - generating a second inference result on the basis of the first inference result and the inference conditions information for controlling a radio access function on the basis of the second inference result.

6. The apparatus of claim 5, wherein the means are further configured for generating the second inference result additionally on the basis of one or more further inference results and associated inference conditions information from one or more further wireless devices.

7. The apparatus of claim 5 or 6, wherein the means are further configured for evaluating accuracy or reliability of the first inference result on the basis of the inference conditions information and further information available to the second wireless device.

8. The apparatus of any preceding claim 5 to 7, wherein the means are further configured for controlling the radio access function on the basis of the second inference result.

9. The apparatus of any preceding claim 5 to 8, wherein the means are further configured for transmitting a control message to the first wireless device, indicative of one or more types of inference conditions information to be provided for the second wireless device.

10. The apparatus of claim 2 or 9, wherein the control message comprises an indication of an inference conditions reporting type, and the inference conditions information is included in a report in accordance with the inference reporting type.

11. The apparatus of any preceding claim 5 to 10, wherein the means are further configured for transmitting an inference result reporting configuration to the first device for controlling a report comprising the first inference result.

12. The apparatus of any preceding claim, wherein the inference conditions information comprise radio access parameter information not known by, or otherwise being reported to the second wireless device, before receiving a report comprising the inference conditions information.

13. The apparatus of any preceding claim, wherein the inference conditions information classifies or characterizes at least one of

- side information not associated with the prediction algorithm available to the first wireless device,

- the prediction algorithm, and

- an input to the prediction algorithm.

14. The apparatus of any preceding claim, wherein the inference conditions information is indicative of at least one type of input to the prediction algorithm.

15. The apparatus of any preceding claim, wherein the inference conditions information is included in an inference report comprising the first inference result or wherein the inference conditions information is transmitted in a dedicated inference conditions information report separate from the inference report.

16. The apparatus of any preceding claim, wherein the radio access control function comprises a handover or beam management function and the inference results comprise a radio cell handover prediction or a beam selection prediction.

17. The apparatus of any preceding claim, wherein the inference condition information is transmitted as a part of a handover request.

18. The apparatus of claim 14, wherein the inference conditions information indicates if the prediction algorithm was input by one or more of: signal to noise inference ratio to handover source cell and handover target cell, interference level from other cells, movement velocity considered or not considered, exploration active or not active, best beam identifier, antenna port signal to noise ratio, inter cell cross talk, sequence of applied antenna panels.

19. The apparatus of any preceding claim, wherein the inference conditions information indicates one or more of: validation accuracy of the first inference result, inference operating mode, all inference values obtained from the prediction algorithm included or not included, cause output information indicative of identified one or more causes to obtaining the first inference result, machine- learning model or architecture information, and importance of one or more included conditions.

20. The apparatus of any preceding claim, wherein the inference conditions information is included in one or more of a radio resource control complete message, a radio resource control disconnect message, and a radio resource control condition report message.

21. The apparatus of any preceding claim, wherein the inference or an output of the prediction algorithm comprises identification of an event.

22. The apparatus of any preceding claim, wherein the means comprise at least one processor; and at least one memory including computer program code, the at least one memory and computer program code configured to, with the at least one processor, cause the performance of the apparatus.

23. A method for a first wireless device, comprising:

- determining at least one parameter of a prediction algorithm;

24. The method of claim 23, further comprising: receiving a control message from the second wireless device, indicative of one or more types of inference conditions information to be provided for the second wireless device, and obtaining the inference conditions information in the first wireless device in accordance with the received control message.

25. The method of claim 23, further comprising: determining the inference condition information based on predefined one or more types of inference condition information, or based on determining a new type of inference condition information.

26. The method of any preceding claim, further comprising: receiving an inference result reporting configuration from the second device and generating a report comprising at least the first inference result in accordance with the received inference result reporting configuration.

27. A method for a second wireless device, comprising:

- receiving from a first wireless device inference conditions information associated with a first inference result and characterizing conditions under which the first inference result was obtained by a prediction algorithm in the first wireless device, and

- generating a second inference result on the basis of the first inference result and the inference conditions information for controlling a radio access function on the basis of the second inference result.

28. The method of claim 27, wherein the second inference result is generated additionally on the basis of one or more further inference results and associated inference conditions information from one or more further wireless devices.

29. The method of claim 27 or 28, further comprising: evaluating accuracy or reliability of the first inference result on the basis of the inference conditions information and further information available to the second wireless device.

30. The method of any preceding claim 27 to 29, further comprising controlling the radio access function on the basis of the second inference result.

31. The method of any preceding claim 27 to 30, further comprising transmitting a control message to the first wireless device, indicative of one or more types of inference conditions information to be provided for the second wireless device.

32. The method of claim 24 or 31, wherein the control message comprises an indication of an inference conditions reporting type, and the inference conditions information is included in a report in accordance with the inference reporting type.

33. The method of any preceding claim 27 to 32, further comprising transmitting an inference result reporting configuration to the first device for controlling a report comprising the first inference result.

34. The method of any preceding claim, wherein the inference conditions information comprise radio access parameter information not known by, or otherwise being reported to the second wireless device, before receiving a report comprising the inference conditions information.

35. The method of any preceding claim, wherein the inference conditions information classifies or characterizes at least one of

- the prediction algorithm, and

- an input to the prediction algorithm.

36. The method of any preceding claim, wherein the inference conditions information is indicative of at least one type of input to the prediction algorithm.

37. The method of any preceding claim, wherein the inference conditions information is included in an inference report comprising the first inference result or wherein the inference conditions information is transmitted in a dedicated inference conditions information report separate from the inference report.

38. The method of any preceding claim, wherein the radio access control function comprises a handover or beam management function and the inference results comprise a radio cell handover prediction or a beam selection prediction.

39. The method of any preceding claim, wherein the inference condition information is transmitted as a part of a handover request.

40. The method of any preceding claim, wherein the inference conditions information indicates if the prediction algorithm was input by one or more of: signal to noise inference ratio to handover source cell and handover target cell, interference level from other cells, movement velocity considered or not considered, exploration active or not active, best beam identifier, antenna port signal to noise ratio, inter cell cross talk, sequence of applied antenna panel

41. The method of any preceding claim, wherein the inference conditions information indicates one or more of: validation accuracy of the first inference result, inference operating mode, all inference values obtained from the prediction algorithm included or not included, cause output information indicative of identified one or more causes to obtaining the first inference result, machine- learning model or architecture information, and importance of one or more included conditions.

42. The method of any preceding claim, wherein the inference conditions information is included in one or more of a radio resource control complete message, a radio resource control disconnect message, and a radio resource control condition report message.

43. The method of any preceding claim, wherein the inference of an output of the prediction algorithm comprises identification of an event.

44. A computer program, comprising instructions for causing an apparatus to perform the method of any one of claims 22 to 43.

45. A non-transitory computer readable medium comprising program instructions for causing an apparatus to perform the method of any one of claims 22 to 43.