Channel prediction performance monitoring

Abstract

A method, apparatus, and computer program are described comprising: determining, by a terminal device, first predictions of channel state information using a first predictor; determining, by the terminal device, reference predictions of channel state information using a reference predictor; determining, by the terminal device, estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions; determining, by the terminal device, first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining, by the terminal device, reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

Description

[0001] CHANNEL PREDICTION PERFORMANCE MONITORING

[0002] Field

[0003] Example embodiments may relate to terminal devices, networks, network nodes, and methods for monitoring errors in channel prediction.

[0004] Background

[0005] The radio channel between a base station and terminal device (such as a user equipment) may vary over time in an at least partly predictable manner. For example, a radio channel between a moving terminal device and a stationary base station may vary over time in a manner that is at least partly predictable. In communications between terminal devices and radio access networks, performance may be improved by taking properties of the radio channel into account when performing the communications, such as when determining a suitable direction for beamforming. Variations in the properties of the radio channel between a measurement of the channel and performing communications based on that measurement may reduce performance. There may therefore be an interest in predicting properties of a radio channel and monitoring the performance of the channel predictions.

[0006] Summary

[0007] 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.

[0008] A first aspect provides a terminal device comprising: means for determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; means for determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; means for determining estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions; means for determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and means for determining reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to therespective prediction occasions.

[0009] In some example embodiments, the first predictions are determined using the first predictor based on measurements of the communication channel performed during respective observation windows that precede the respective prediction occasions; and the reference predictions are determined using the reference predictor based on measurements of the communication channel performed during respective observation windows that precede the respective prediction occasions.

[0010] In some example embodiments, the first predictor comprises an artificial intelligence, Al, and / or machine learning, ML, model. In some example embodiments, the reference predictor comprises a Kalman filter or a zero-order hold filter.

[0011] In some example embodiments, the terminal device further comprises means for sending information derived from the first prediction errors and the reference prediction errors to a network node. In some example embodiments, the information comprises the first prediction errors and the reference prediction errors.

[0012] In some example embodiments, the terminal device further comprises means for determining a prediction performance indicator by characterising a statistical distribution of the first prediction errors with respect to a statistical distribution of the reference prediction errors.

[0013] In some example embodiments, the means for determining the prediction performance indicator is configured to determine the prediction performance indicator based at least in part on a size of an area overlap and / or a size of an area of non-overlap between the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors. In some example embodiments, the means for determining the prediction performance indicator is configured to determine the prediction performance indicator based at least in part on one or more of: a comparison of a mean of the statistical distribution of the first prediction errors and a mean of the statistical distribution of the reference prediction errors; and a comparison of a variance or standard deviation of the statistical distribution of the first prediction errors and a variance or standard deviation of the statistical distribution of the reference prediction errors.

[0014] In some example embodiments, the terminal device further comprises means for determining a correction factor based on at least one of the following: at least one measurement of the communication channel; a doppler spread of the communication channel; an indication of a noise level of the communication channel; and an indication of an interference level of the communication channel. In some example embodiments, theterminal device further comprises means for applying the correction factor to the prediction performance indicator.

[0015] In some example embodiments, the means for determining the correction factor is configured to determine the correction factor using an AI / ML model.

[0016] In some example embodiments, the terminal device further comprises means for receiving from a network node an indication of a correction factor upper and / or lower limit, and the means for determining a correction factor is configured to determine the correction factor based on said limit.

[0017] In some example embodiments, the information comprises the prediction performance indicator.

[0018] In some example embodiments, the terminal device further comprises: means for comparing the prediction performance indicator to a threshold value; and means for sending an indication of the result of the comparison to a network node. In some example embodiments, the terminal device further comprises: means for comparing the predictor performance indicator to a threshold value; and means for determining whether to use the first predictor or a different predictor based on a result of the comparison. In some example embodiments, the terminal device further comprises means for receiving an indication of the threshold value from the network.

[0019] In some example embodiments, the terminal device further comprises means for receiving, from a network node, an indication of a number of first prediction errors and reference prediction errors from which to derive the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

[0020] In some example embodiments, the terminal device further comprises: means for determining second predictions of channel state information using a second predictor, wherein the second predictions correspond to predictions of channel state information of a communication channel at the respective prediction occasions; and means for determining second prediction errors based on respective comparisons of the estimates of the channel state information and the second predictions of the channel state information corresponding to the respective prediction occasions.

[0021] In some example embodiments, the terminal device further comprises: means for determining a second prediction performance indicator by characterising a statistical distribution of the second prediction errorswith respect to a statistical distribution of the reference prediction errors; and means for comparing the first and second prediction performance indicators.

[0022] In some example embodiments, the terminal device further comprises means for selecting the reference predictor.

[0023] A second aspect provides a method comprising: determining, by a terminal device, first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining, by the terminal device, reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining, by the terminal device, estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions; determining, by the terminal device, first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining, by the terminal device, reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0024] In some example embodiments, the first predictions are determined using the first predictor based on measurements of the communication channel performed during respective observation windows that precede the respective prediction occasions; and the reference predictions are determined using the reference predictor based on measurements of the communication channel performed during respective observation windows that precede the respective prediction occasions.

[0025] In some example embodiments, the first predictor comprises an artificial intelligence, Al, and / or machine learning, ML, model.

[0026] In some example embodiments, the reference predictor comprises a Kalman filter or a zero-order hold filter.

[0027] In some example embodiments, the method further comprises sending, from a terminal device, information derived from the first prediction errors and the reference prediction errors to a network node.In some example embodiments, the information comprises the first prediction errors and the reference prediction errors.

[0028] In some example embodiments, the method further comprises: determining, by the terminal device, a prediction performance indicator by characterising a statistical distribution of the first prediction errors with respect to a statistical distribution of the reference prediction errors.

[0029] In some example embodiments, determining the prediction performance indicator is based at least in part on a size of an area overlap and / or a size of an area of non-overlap between the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

[0030] In some example embodiments, determining the prediction performance indicator is based at least in part on one or more of: a comparison of a mean of the statistical distribution of the first prediction errors and a mean of the statistical distribution of the reference prediction errors; and a comparison of a variance or standard deviation of the statistical distribution of the first prediction errors and a variance or standard deviation of the statistical distribution of the reference prediction errors.

[0031] In some example embodiments, the method further comprises determining, by the terminal device, a correction factor based on at least one of the following: at least one measurement of the communication channel; a doppler spread of the communication channel; an indication of a noise level of the communication channel; and an indication of an interference level of the communication channel. In some example embodiments, the method further comprises applying, by the terminal device, the correction factor to the prediction performance indicator.

[0032] In some example embodiments, determining the correction factor comprises determining the correction factor using an AI / ML model.

[0033] In some example embodiments, the information comprises the prediction performance indicator.

[0034] In some example embodiments, the method further comprises receiving, at the terminal device, from a network node, an indication of a correction factor upper and / or lower limit, and the determining of the correction factor is based on said limit.

[0035] In some example embodiments, the method further comprises: comparing, by the terminal device, theprediction performance indicator to a threshold value; and sending, form the terminal device, an indication of the result of the comparison to a network node.

[0036] In some example embodiments, the method further comprises: comparing, by the terminal device, the predictor performance indicator to a threshold value; and determining, by the terminal device, whether to use the first predictor or a different predictor based on a result of the comparison.

[0037] In some example embodiments, the method further comprises receiving, at the terminal device, an indication of the threshold value from the network.

[0038] In some example embodiments, the method further comprises receiving, at the terminal device, from a network node, an indication of a number of first prediction errors and reference prediction errors from which to derive the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

[0039] In some example embodiments, the method further comprises: determining, by the terminal device, second predictions of channel state information using a second predictor, wherein the second predictions correspond to predictions of channel state information of a communication channel at the respective prediction occasions; and determining, by the terminal device, second prediction errors based on respective comparisons of the estimates of the channel state information and the second predictions of the channel state information corresponding to the respective prediction occasions.

[0040] In some example embodiments, the method further comprises: determining, by the terminal device, a second prediction performance indicator by characterising a statistical distribution of the second prediction errors with respect to a statistical distribution of the reference prediction errors; and comparing, by the terminal device the first and second prediction performance indicators.

[0041] In some example embodiments, the method further comprises selecting, by the terminal device, the reference predictor.

[0042] A third aspect provides a computer program comprising instructions which, when executed by at least one processor, cause an apparatus to perform: determining, by the apparatus, first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining, by the apparatus,reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining, by the apparatus, estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions; determining, by the apparatus, first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining, by the apparatus, reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0043] In some example embodiments, the third aspect may include any other feature mentioned with respect to the method of the second aspect.

[0044] A fourth aspect provides a non-transitory computer-readable medium having instructions stored thereon which, when executed by at least one processor, cause an apparatus to perform: determining, by the apparatus, first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining, by the apparatus, reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining, by the apparatus, estimates of channel state information of the communication channel at the respective prediction occasions based at least in parton respective measurements of the communication channel at the respective prediction occasions; determining, by the apparatus, first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining, by the apparatus, reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0045] In some example embodiments, the fourth aspect may include any other feature mentioned with respect to the method of the second aspect.

[0046] A fifth aspect provides an apparatus, the apparatus having at least one processor and at least one memory having instructions stored thereon which, when executed by the at least one processor, cause the apparatusto perform: determining, by the apparatus, first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining, by the apparatus, reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining, by the apparatus, estimates of channel state information of the communication channel at the respective prediction occasions based at least in parton respective measurements of the communication channel at the respective prediction occasions; determining, by the apparatus, first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining, by the apparatus, reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0047] In some example embodiments, the fifth aspect may include any other feature mentioned with respect to the method of the second aspect.

[0048] A sixth aspect provides a network node comprising means for sending, to a terminal device, a configuration configuring the terminal device to perform a method comprising: determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining estimates of channel state information at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions; determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0049] In some example embodiments the network node further comprises means for receiving, from the terminal device, information derived from the first prediction errors and the reference prediction errors.In some example embodiments the information comprises the first prediction errors and the reference prediction errors.

[0050] In some example embodiments the network node further comprises means for determining a prediction performance indicator by characterising a statistical distribution of the first prediction errors with respect to a statistical distribution of the reference prediction errors.

[0051] In some example embodiments the information comprises a prediction performance indicator.

[0052] In some example embodiments the network node further comprises means for comparing the prediction performance indicator to a threshold value.

[0053] In some example embodiments the network node further comprises means for receiving, from the terminal device, an indication of the result of a comparison between the prediction performance indicator and a threshold value.

[0054] In some example embodiments the network node further comprises means for sending, to the terminal device, an indication of the threshold value.

[0055] In some example embodiments the network node further comprises means for sending, to the terminal device, an indication of a number of first prediction errors and reference prediction errors from which to derive the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

[0056] In some example embodiments the network node further comprises means for sending, to the terminal device, an indication of a correction factor upper and / or lower limit.

[0057] A seventh aspect provides a method comprising: sending, from a network node, to a terminal device, a configuration configuring the terminal device to perform a method comprising: determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining estimates of channel state information of the communicationchannel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions; determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0058] In some example embodiments, the method further comprises receiving, at the network node, from the terminal device, information derived from the first prediction errors and the reference prediction errors.

[0059] In some example embodiments, the information comprises the first prediction errors and the reference prediction errors.

[0060] In some example embodiments the method further comprises: determining, by the network node, a prediction performance indicator by characterising a statistical distribution of the first prediction errors with respect to a statistical distribution of the reference prediction errors.

[0061] In some example embodiments the information comprises a prediction performance indicator.

[0062] In some example embodiments the method further comprises comparing, by the network node, the prediction performance indicator to a threshold value.

[0063] In some example embodiments the method further comprises receiving, at the network node, from the terminal device, an indication of the result of a comparison between the prediction performance indicator and a threshold value.

[0064] In some example embodiments the method further comprises sending, from the network node, to the terminal device, an indication of the threshold value.

[0065] In some example embodiments the method further comprises sending, from the network node, to the terminal device, an indication of a number of first prediction errors and reference prediction errors from which to derive the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.In some example embodiments the method further comprises sending, to the terminal device, an indication of a correction factor upper and / or lower limit.

[0066] An eighth aspect provides a computer program comprising instructions which, when executed by at least one processor, cause an apparatus to perform: sending, from the apparatus, to a terminal device, a configuration configuring the terminal device to perform a method comprising: determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions; determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0067] In some example embodiments, the eighth aspect may include any other feature mentioned with respect to the method of the seventh aspect.

[0068] A ninth aspect provides a non-transitory computer-readable medium having instructions stored thereon which, when executed by at least one processor, cause an apparatus to perform: sending, from the apparatus, to a terminal device, a configuration configuring the terminal device to perform a method comprising: determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions; determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining reference prediction errors based on respective comparisons of the estimates of the channelstate information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0069] In some example embodiments, the ninth aspect may include any other feature mentioned with respect to the method of the seventh aspect.

[0070] A tenth aspect provides an apparatus, the apparatus having at least one processor and at least one memory having instructions stored thereon which, when executed by the at least one processor, cause the apparatus to perform: sending, from the apparatus, to a terminal device, a configuration configuring the terminal device to perform a method comprising: determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions; determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions; determining estimates of channel state information of the communication channel at the respective prediction occasions based at least in parton respective measurements of the communication channel at the respective prediction occasions; determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; and determining reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

[0071] In some example embodiments, the tenth aspect may include any other feature mentioned with respect to the method of the seventh aspect.

[0072] Brief Description of the Drawings

[0073] Example embodiments will now be described by way of non-limiting example, with reference to the accompanying drawings, in which:

[0074] Fig. 1 illustrates an example of a communication system to which examples disclosed herein may be applied;

[0075] Fig. 2 illustrates aspects of this disclosure schematically;

[0076] Figs. 3a - 3c illustrate examples of reference prediction error distributions and first prediction errordistributions in accordance with example embodiments;

[0077] Fig. 4 is a block diagram illustrating a system in accordance with example embodiments;

[0078] Fig. 5 is a message flow diagram illustrating a method in accordance with example embodiments;

[0079] Figs. 6 and 7 are flow diagrams illustrating methods in accordance with example embodiments;

[0080] Fig. 8 is a schematic diagram of a system that may be used to implement one or more of the example embodiments; and

[0081] Fig. 9 shows tangible media for storing computer-readable code which when run by a computer may perform methods according to example embodiments described herein.

[0082] Detailed Description

[0083] The following embodiments are exemplary. Although the specification may refer to “an”, “one”, or “some” embodiment(s) in several locations of the text, this does not necessarily mean that each reference is made to the same embodiment(s), or that a particular feature only applies to a single embodiment. Single features of different embodiments may also be combined to provide other embodiments. Further, when a particular feature, structure, or characteristic is described in connection of an embodiment, it is within the knowledge of one skilled in the art to apply such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described. It shall be understood that although the terms “first,” “second” and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another.

[0084] For the purposes of the present disclosure, the phrases “at least one of A or B”, “at least one of A and B”, and “A and / or B” means (A), (B), or (A and B). For the purposes of the present disclosure, the phrase “A, B, and / or C” means (A), (B), (C), (A and B), (A and C), (B and C), or (A, B, and C).

[0085] Embodiments described may be implemented in a communication system, such as any of the following radio access technologies (RATs): World-wide Interoperability for Micro-wave Access (WiMAX), Global System for Mobile communications (GSM, 2G), GSM EDGE radio access Network (GERAN), General Packet Radio Service (GRPS), Universal Mobile Telecommunication System (UMTS, 3G) based on basic wideband-code division multiple access (W-CDMA), high-speed packet access (HSPA), Long Term Evolution (LTE), LTE-Advanced, and enhanced LTE (eLTE), 5G (also called NR), or any future RAT such as 6G. Moreover, communication within the communication system may utilize any proper wireless communication technology, comprising but not limited to: Code Division Multiple Access (CDMA), Frequency Division Multiple Access (FDMA), Time Division Multiple Access (TDMA), Frequency Division Duplexing (FDD), Time DivisionDuplexing (TDD), Multiple-Input Multiple-Output (MIMO), Orthogonal Frequency Division Multiplexing (OFDM), and / or Discrete Fourier Transform spread OFDM (DFT-s-OFDM).

[0086] As used herein, the term “network device” or “network node” refers to a node in a communication system via which user equipment may access the network and / or which is capable of controlling radio communication and managing radio resources within a cell. The network node or network device may be referred to as a base station (BS), an access point (AP) or an access node. The network device may be, depending on the applied technology, for example, a node B (NodeB or NB), an evolved NodeB (eNodeB or eNB), an NR NB (also referred to as a gNB), a Remote Radio Unit (RRU), a radio head (RH), a remote radio head (RRH), a relay, an Integrated Access and Backhaul (IAB) node, a low power node, a non-terrestrial network (NTN) node or non-ground network device such as a satellite network device, a low earth orbit (LEO) satellite and a geosynchronous earth orbit (GEO) satellite, or an aircraft network device.

[0087] Moreover, in connection of split radio access network (RAN), the network device may refer to a centralised unit (CU) of a base station and / or a distributed unit (DU) of a base station. An interface between CU and DU may be referred to as an F1 interface in NR. In the split RAN architecture, node operations may be carried out, at least partly, in the central / centralized unit, CU, (e.g. server, host or node) operationally coupled to the DU, (e.g. a radio head / node). One CU may control one or more DUs, acting at least as transmit / receive (Tx / Rx) nodes. In some embodiments, the DUs may comprise e.g. a radio link control (RLC), medium access control (MAC) layer and a physical (PHY) layer, whereas the CU may comprise the layers above RLC layer, such as a packet data convergence protocol (PDCP) layer, Service Data Application Protocol (SDAP) layer and a radio resource control (RRC) layer. Other functional splits are possible too. In practice, any processing task may be performed in either the CU and / or the DU and the boundary where the responsibility is shifted between the CU and the DU may depend on the applied implementation.

[0088] The term “terminal device” refers to any end device that may be capable of wireless communication. By way of example, a terminal device may be referred to as a communication device, user equipment (UE), a Subscriber Station (SS), or a Mobile Station (MS). The terminal device may include a mobile phone, a cellular phone, a smart phone, voice over IP (VoIP) phones, wireless local loop phones, a tablet, a wearable terminal device, a personal digital assistant (PDA), portable computers, desktop computer, image capture terminal devices such as digital cameras, gaming terminal devices, music storage and play-back appliances, vehiclemounted wireless terminal devices, USB dongles, an Internet of Things (loT) device, a watch or other wearable, a head-mounted display (HMD), a vehicle, a drone, a medical device and applications (e.g., remote surgery), an industrial device and applications (e.g., a robot and / or other wireless devices operating in anindustrial and / or an automated processing chain contexts), a consumer electronics device, a device operating on commercial and / or industrial wireless networks, and the like.

[0089] A term “resource”, as used herein, may refer to radio resources in time domain, in frequency domain, in space domain, and / or in code domain. Some examples of resources include e.g. a physical resource block (PRB), a radio frame, a subframe, a time slot, a subband, a frequency region, a sub-carrier, a beam, etc. The term “transmission” and / or “reception” may refer to wirelessly transmitting and / or receiving via a wireless propagation channel on radio resources.

[0090] Fig. 1 illustrates an example of a communication system to which examples disclosed herein may be applied. The communication system or a cellular communication system may comprise a network node 110 providing one or more cells, such as cell 100, and a network node 112 providing one or more other cells, such as cell 102. Each cell may be, e.g., a macro cell, a micro cell, femto, or a pico cell, for example. The cell may define a coverage area or a service area of the corresponding access node.

[0091] The network node 110 may provide a user equipment (UE) 120 (one or more UEs) with wireless access to the communication system. The wireless access may comprise downlink (DL) communication from the network node 110 to the UE 120 and uplink (UL) communication from the UE 120 to the network node 110. Examples of uplink channels comprise physical uplink control channel (PUCCH) for transmitting control information and physical uplink shared channel (PUSCH) for transmitting data towards the network. Examples of downlink channels comprise physical downlink control channel (PDCCH) for transmitting control information and physical downlink shared channel (PDSCH) for transmitting data towards the user equipment.

[0092] There may be a plurality of UEs 120, 122 in the system. Each of them may be served by the same or by different network nodes 110, 112. UE may be configured with dual connectivity (DC), wherein the UE, e.g. UE 120, may be connected to multiple network nodes 110, 112. The UEs 120, 122 may communicate with each other, in case device-to-device (D2D) communication interface is established between them via a so-called sidelink (SL). Such D2D communications may be referred to as machine-to-machine, peer-to-peer (P2P) communications, or vehicle-to-vehicle (V2V), for example.

[0093] In the case of multiple network nodes in the communication system, the network nodes may be connected to each other via an interface. LTE specifications call such an interface as X2 interface. An interface between an LTE node and a 5G node, or between two 5G nodes is called Xn interface.The network nodes 110 and 112 may be further connected via another interface to a core network 116 of the communication system. The LTE specifications specify the core network as an evolved packet core (EPC), and the core network may comprise e.g. a mobility management entity (MME) and a gateway node. The MME may handle mobility of terminal devices in a tracking area encompassing a plurality of cells and handle signalling connections between the terminal devices and the core network. The gateway node may handle data routing in the core network and to / from the terminal devices. The 5G specifications specify the core network as a 5G core (5GC). The 5G core may comprise e.g. an access and mobility management function (AMF) and a user plane function / gateway (UPF) and other functions. The AMF may handle termination of non-access stratum (NAS) signalling, NAS ciphering & integrity protection, registration management, connection management, mobility management, access authentication and authorization, security context management. The UPF node may support packet routing and forwarding, packet inspection and quality of service (QoS) handling, for example.

[0094] Channel Prediction

[0095] In communications between a radio access network and a terminal device over a radio channel, it may be possible to improve communications between the terminal device and network by accounting for properties of the radio channel. For example, it may be possible to enhance signal reception by selecting an appropriate beamforming direction.

[0096] The radio channel may be estimated, for example, by performing measurements on reference signals. For example, downlink reference signals such as channel state information reference signals (CSI-RS) may be measured to probe the properties of the radio channel.

[0097] However, performing an estimation of the channel properties based on measurements and configuring a network or terminal device to account for estimated properties of the channel may take time. In this time, the channel properties may evolve. In addition, opportunities to measure the channel do not necessarily occur immediately before a data communication over the channel is scheduled. The time evolution of channel properties may therefore reduce communication performance.

[0098] Prediction of channel properties (i.e. , channel state information, or CSI) may allow a network or terminal device to be configured to account for predicted properties of the channel at the time that a communication occurs. A channel predictor, for predicting CSI may therefore be used by a terminal device or network deviceto improve communications.

[0099] A channel predictor may use data such as past measurements, or past estimates, of a radio channel to predict future CSI.

[0100] Different approaches may be taken when designing a channel predictor. For example, a channel predictor may be designed based on a set of assumptions (e.g., a Kalman filter may make predictions based on an assumed state transition model).

[0101] One approach to designing a channel predictor is to use artificial intelligence (Al) / machine learning techniques (ML), such as deep learning, to perform channel prediction. For example, channel measurements or estimates may be used as input data, and CSI at a subsequent time may be the output.

[0102] Regardless of which method of performing channel prediction is used, the effectiveness of the channel predictor may vary depending on the input data. For example, in the former case, the assumptions underlying the predictive model may not hold in all circumstances, and in the latter case the training data on which the ML channel predictor was trained may have been incomplete or obsolete (i.e. , not have included training data relevant to some encountered input data). Depending on the performance of the predictor, using time and computational resources to perform prediction may be unjustified, and in the worst case, the predictions may be so erroneous that relying on them would reduce performance.

[0103] Performance metrics that quantify the performance of a channel predictor may assist a terminal device or a network device in assessing whether a channel predictor should be used, or if multiple channel predictors are available, assessing which channel predictor should be used.

[0104] One option for assessing the performance of a channel predictor is to quantify the difference between predicted CSI and measured CSI. For example, squared generalized cosine similarity (SGCS) may be used to indicate the alignment of a predicted vector with a vector later estimated from measurements, or normalized mean squared error (NMSE) may be used to quantify the error of a predicted channel transfer function with respect to a channel transfer function later estimated from measurements.

[0105] However, even if SGCS or NMSE are calculated, these do not (in isolation) indicate the extent to which using a channel predictor improves performance. In poor radio conditions (e.g., if the channel coherence time is low, the channel has a large delay spread, the channel is noisy, or if interference is present), predicted CSImay differ significantly from the ground truth, but predicted CSI may still be an improvement over alternatives (such as performing no prediction). Even a perfectly performing channel predictor (e.g., that accounts for all of the behaviors of the channel that are predictable) may produce very different performance metrics depending on signal to interference plus noise (SINR), terminal device velocity, frequency selectivity, or other channel parameters. Metrics such as SGCS and NMSE do not separate channel predictor performance from the effect of channel conditions, so evaluating these metrics against a fixed threshold may result in an underperforming predictor appearing effective in favorable channel conditions, and an effective predictor appearing to be ineffective in unfavorable channel conditions.

[0106] This might lead to poor decisions, such as the activation of a less suitable predictor, the premature or late fallback to a more robust predictor (e.g., a predictor that is less sensitive to channel conditions), or the use of no prediction method, where prediction would still provide performance gains.

[0107] A large number of factors unrelated to the channel predictor may affect the performance of the channel predictor, so it may be impractical to define separate SGCS or NMSE thresholds based on all of the factors that could affect prediction performance.

[0108] Different instances of prediction may produce different errors even under the same channel conditions (e.g., because of noise in the estimates of CSI, or noise in the predicted CSI). To obtain more complete information on the performance of the channel predictor, errors at a number of different instances may be determined, and the statistics of these errors may be considered. For example, multiple time, frequency, and / or space instances may be considered.

[0109] Relative performance monitoring

[0110] Aspects disclosed herein relate to determining the performance a channel predictor relative to another channel predictor (e.g., a simpler channel predictor). A performance threshold may be defined in terms of this relative performance.

[0111] Fig. 2 illustrates an aspect of this disclosure schematically, on timeline 200.

[0112] Timeline 200 shows a set of estimates 210 of CSI during an observation window, a prediction 220 of CSI at a prediction occasion 232 and determined using a first predictor, a prediction 230 of CSI at the prediction occasion 232 and determined using a reference predictor, and a channel estimate 240 of CSI at the predictionoccasion 232. A prediction occasion may correspond to a time instance or period at which or during which the predicted CSI may be assumed to be valid. For example, the prediction occasion may correspond to a mini-slot, slot, sub-frame, or transmission time interval (TTI).

[0113] To simplify this example, the CSI is shown on timeline 200 as a dimensionless scalar. In some examples, the predicted CSI may be a channel transfer function, a beam or precoding vector, a beam ID, etc.

[0114] The set of estimates 210 may be determined from measurements during an observation window 212. A window 222 for determining the predictions using the predictors may be provided before the prediction occasion 232. The prediction occasion 232 may correspond to the time instance for which the predictors attempt to predict the CSI, based on estimates of the CSI during the observation window 212.

[0115] Measurements of CSI may be made using scheduled reference signals. When determining the performance of the predictor, a prediction occasion may therefore be selected to correspond to a scheduled reference signal, so that the prediction may be compared with a corresponding measurement at the prediction occasion. Therefore, an estimate 240 of the channel at the prediction occasion 232 is determined. A measurement on which this estimate is based may take place at the prediction occasion 232 or slightly before, and the estimate may then be determined.

[0116] The reference predictor may be a simpler predictor than the first predictor. In some examples, the reference predictor takes the most recently measured CSI (in the observation window) to be the predicted CSI. Such a reference predictor may be referred to as a zero-order hold predictor.

[0117] The first predictor may be an AI / ML based predictor (e.g., comprising a deep learning model trained on example CSI data), or some other predictor. This predictor may use estimated CSI in the observation window as input, and output a prediction.

[0118] In the simple case shown on timeline 200, an error of the first prediction may correspond to the difference between prediction 220 and estimate 240, while an error of the reference prediction may correspond to the difference between prediction 230 and estimate 240.

[0119] In the case of more complicated CSI, a different error metric may be selected. For example, SGCS may provide an error metric for a vector (with a higher value corresponding to a lower error / better alignment between vectors), or NMSE may provide an error metric for a channel transfer function (with a lower valuecorresponding to a lower error).

[0120] Timeline 200 shows one set of predictions, corresponding to respective first and reference prediction errors. This process may be repeated for a different (e.g., subsequent) prediction occasion. The observation window may move (e.g., to encompass additional estimates), and, for later prediction occasions, estimate 240 may fall within the observation window and be used for predictions.

[0121] Repeating this process to obtain a set of errors may allow for the statistical distributions of the first and reference prediction errors to be determined, or for the statistics of these distributions to be determined.

[0122] Figs. 3a, 3b, and 3c illustrate examples of reference prediction error distribution 320 and first prediction error distribution 330. The below discussion is applicable to at least distributions in the form of normalized histograms, unnormalized histograms, and continuous probability density functions, so the vertical axis has not been labelled, but could correspond to a probability, probability density, number of observations, etc., as appropriate.

[0123] The prediction errors in this example are quantified by SGCS (so a higher value corresponds to a lower error / better alignment between predicted and estimated vectors). This discussion is applicable to other measures of error (such as NMSE, where a lower value corresponds to a lower error).

[0124] To define a measure of relative performance, first prediction error distribution 330 may be compared with, or characterized with respect to, reference prediction error distribution 320. Several metrics may be extracted from the errors or the distribution of errors and used in combination or in isolation to determine a relative performance metric.

[0125] In Fig. 3a, first prediction error distribution 330 has a higher mean SGCS (corresponding to a better prediction) than reference prediction error distribution 320, a lower variance, and no overlap with reference prediction error distribution.

[0126] In Fig. 3b, first prediction error distribution 330 has a higher mean SGCS than reference prediction error distribution 320 and a lower variance than reference prediction error distribution 320 but has an area of overlap 334 with reference prediction error distribution 320.

[0127] In Fig. 3c, first prediction error distribution 330 has a lower mean SGCS than reference prediction errordistribution 320 and a lower variance than reference prediction error distribution 320, and has an area of overlap 334 with reference prediction error distribution 320.

[0128] The first predictor of each of these figures may be the same (e.g., the channel conditions in which the error data is collected may differ between Figs. 3a - 3c, and the first predictor may be better suited to some channel conditions) or the first predictor of each of these figures may be different (e.g., the channel conditions could be the same in each figure, and the different predictors may perform differently because they are differently suited to those channel conditions).

[0129] In the former case, metrics derived from Figs. 3a - 3c may be used to determine whether to use the predictor (e.g., a device may determine from a derived metric whether the performance gain from using the first predictor is worth the cost in different channel conditions). In the latter case, metrics derived from Figs. 3a -3c may be used to determine which predictor to use in the current channel conditions, if any.

[0130] A relative metric may consider the distance between the means (e.g., by subtracting the mean of distribution 320 from the mean of distribution 330, a metric with a higher value corresponding to better relative performance may be obtained). According to this metric, the first distributions of Figs. 3a and 3b are better than the reference distributions, but the first distribution of Fig. 3a is better by a greater amount. For Fig. 3c, this metric may be negative (or the metric could be bounded at zero, to represent no improvement).

[0131] The relative metric may additionally or alternatively account for variance of one or both of the distributions. In one example, the distance between the means may be divided by the variance (or standard deviation) of one (or both) of the distributions, so that the metric may take the statistical significance of an improvement over the reference distribution into account. For example, when adjusting the difference between the means based on the variance of the first distribution (e.g., by dividing the difference by the variance), the lower variance of the first distribution of Fig. 3a compared to the first distribution of Fig. 3b may increase the difference between the relative metrics for Fig. 3a and Fig. 3b.

[0132] A relative metric may account for the size of areas of overlap or non-overlap of the distributions. The total area under each distribution may be normalized, to assist in comparison of the distributions if the areas are not already equal, or to assist in comparison between metrics derived from the area.

[0133] In some examples, the size of the overlap 334 between the distributions may be used as the relative metric, or the relative metric may be based at least in part on the size of the overlap. The larger overlap 334 in Fig.3b compared to Fig. 3a may suggest that the relative improvement of the first predictor of Fig. 3b is worse than the relative improvement of the first predictor of Fig. 3a. The performance metric may be based on the size of the overlap and the relative means of the first and reference error distributions (e.g., it could be checked that the mean of the first distribution is larger than the mean of the reference error distribution, and the metric could be modified to account for this). This may account for examples such as Fig. 3c, in which the size of the overlap may be similar to Fig. 3b, but the mean of the first error distribution is lower than the mean of the reference error distribution.

[0134] In some examples, the size of one or more areas of non-overlap of the first distribution 330 with the reference distribution 320 may be accounted for in the metric. For example, a large area of non-overlap in the first distribution, such as in the case of Fig. 3a, may indicate significant performance gains, compared to the smaller area of non-overlap in Fig. 3b.

[0135] The relative metric may account for the scenario in Fig. 3c, in which the non-overlap corresponds to worse performance, in different ways. For example, if the mean of the first distribution is lower than the mean of the reference distribution, the metric could be set to a minimum value. In some examples, an area of non-overlap of the first distribution may be divided into the area with an SGCS that is greater than the mean of the reference distribution (e.g. similar to the area of non-overlap of Fig. 3b), and into the area with an SGCS that is less than the mean of the reference distribution (e.g., similar to the area of non-overlap of Fig. 3c), and the area with an SGCS that is less than the mean of the reference distribution may be ignored, or subtracted from the area of non-overlap that is greater than the reference distribution mean when determining the relative performance metric.

[0136] A relative performance metric may be derived from the first and reference distributions in the above described way or in other suitable ways.

[0137] The performance of a simpler predictor may depend on how challenging the channel conditions experienced are (e.g., in terms of noise, etc.), while the performance of a complex predictor, such as an AI / ML predictor, may to a larger extent depend on how well suited the predictor is to the channel conditions (e.g., whether the channel conditions were included in training data when training an AI / ML predictor, or whether the channel conditions were accounted for in the assumptions underlying another predictive model).

[0138] If challenging channel conditions affect both the reference predictor and the first predictor, and both experience a similar drop in performance, then a relative performance metric may be less affected than ametric such as SGCS, while if the channel conditions change to improve the performance of the reference predictor without correspondingly improving the performance of the first predictor, the relative performance metric may indicate this.

[0139] This may allow a failure of the first predictor (e.g., because of a failure of an underlying ML model) to be distinguished from poor channel conditions using the relative performance metric.

[0140] Once the performance metric is determined from first prediction error data and reference prediction error data, a terminal device or network device may compare this performance metric to one or more thresholds. Based on the result of this comparison, a terminal device or network device may determine to send a message, change configuration, or take some other action. For example, the terminal device or network device may determine to change the predictor used, or to not use prediction. In some examples a terminal device or network device may send an indication, instruction, or report to another device based on the result of the comparison. For example, a terminal device may report the relative performance metric to the network based on a comparison performed at the terminal device, and / or a network device may send a terminal device an instruction to use or not use the first predictor based on the comparison. These thresholds may be defined differently for differently defined error statistics. For example, NMSE and SGCS may result in differently shaped error distributions, resulting in differently sized areas of overlap, so relative performance metrics derived from these may have different values for the same degree of predictor performance.

[0141] The distributions 320 and 330 may vary over time (e.g., because channel conditions change), so the metrics may be updated over time. For example, first prediction errors and reference prediction errors may be determined periodically, with older prediction error data being replaced with newer prediction error data (e.g., to increase the recency of the data while maintaining a sample size) to update the distributions periodically.

[0142] Correction factor

[0143] While the above-described metrics may provide a relative measure of the performance of a channel predictor, for some metrics, such as an overlap size, it may be possible to further refine the metric to isolate the performance of the channel predictor from the effects of noisy channel conditions.

[0144] The size of an overlap between a distribution of reference predictor errors and a distribution of first predictor errors may depend at least partly on the breadth of the distributions (which may correspond to the variance) and the positions of the distributions (which may be approximated by the mean of the distributions).Even if the positions of the distributions (e.g., in terms of SGCS, or NMSE, etc.) are unchanged by changing channel conditions, the widths of the distributions may change. For example, the variance of the predictions may depend at least in part on the variance of channel estimates on which the predictions are based. Increasing the widths of the distributions may increase the area of overlap, but if this increase is due to an increase in the variance of the channel estimates, rather than the performance of the channel predictor, this increase in the area of overlap may be misleading.

[0145] Therefore, in some examples, the metric (or a threshold against which it is to be compared) may include a correction factor, which may account (or at least partially account) for effects of the channel conditions on the metric that do not result from the performance of the predictor.

[0146] The correction factor may be determined based on a number of factors. For example, the correction factor may be determined based on a measure of channel noise (e.g., a signal to noise ratio measurement, SNR), delay spread, interference, and or channel coherence time. In some examples, the correction factor may be based on the width, variance, or standard deviation of the first distribution and / or the reference distribution (for example, the area of overlap may be divided by a correction factor linearly increasing with the first or reference error distribution width to obtain a metric).

[0147] In some examples, an AI / ML model, such as a deep neural network, may be trained to determine a correction factor using measurements related to channel conditions. For example, information such as an estimate of the SINR, CSI estimates of a radio channel over multiple time slots, the shape of the error statistics of the reference predictor and / or the first predictor, etc. may be input to the model to infer a correction factor.

[0148] The correction factor (and / or a corresponding corrected performance metric threshold) may have a bounded maximum and / or minimum value. Bounding the correction factor (and / or the thresholds) may avoid undesired behavior in edge cases (e.g., such as the failure of an ML model used to determine a correction factor).

[0149] In some examples, a terminal device may use the relative performance metric in life-cycle management of an ML model. In some examples, the relative performance metric may be used to determine whether to use a potential fallback mode (such as performing no prediction or using a different predictor). Additionally, or alternatively, the terminal device may regularly report the relative performance metric to a network device so that the network can determine whether further action is to be taken.In some examples, more than one reference predictor might be used. For example, in addition to a ZOH predictor, a non-AI / ML predictor and / or multiple different ML model predictors may be used as references. This may increase performance monitoring accuracy at the cost of a higher processing overhead. The relevance of the processing overhead depends then on the general monitoring configuration. In the case of regular monitoring with a relatively high periodicity then the implementation efficiency may be a relevant factor.

[0150] One option is to combine (more) regular performance monitoring using a simpler reference predictor (e.g., a ZOH predictor), with (less) regular performance monitoring using a more complex reference predictor, which may include multiple reference predictors.

[0151] Fig. 4 is a block diagram illustrating a system for determining a relative performance metric according to example embodiments, designated generally by the reference numeral 400. System 400 may be implemented in a terminal device.

[0152] Block 410 of the system is a CSI estimator. This may estimate CSI from channel measurements (e.g., of CSI-RS). Estimated CSI may be used as inputs to reference predictor 412 and first predictor 414 as well as correction factor calculator 420. Predictors 412 and 414 may determine reference predictions and first predictions of CSI at prediction occasions based on estimated CSI corresponding from observation windows preceding the respective prediction occasions.

[0153] The reference predictions and first predictions may respectively be inputs to the reference prediction error statistic estimator 416 and the first prediction error statistic estimator 418. The error statistic estimators may compare the predictions for a particular prediction occasion with an estimate of the CSI at the particular prediction occasion. The CSI for a prediction occasion may be provided by CSI estimator 410. The comparison may produce error statistics for the reference predictions and the first predictions. For example, an SCGS may be estimated based on a comparison of a predicted vector and an estimated vector, or an NMSE value may be based on a comparison of a predicted channel transfer function and an estimated channel transfer function.

[0154] A number of reference and first prediction errors may be collected by the relative performance metric calculator 422, which may compare the distributions of these errors to determine a performance metric. A correction factor, determined by a correction factor calculator 420 may be applied by the performance metric calculator, to account for potential effects of channel conditions on the relative performance metric. In someexamples the correction factor may be determined based (at least in part) on CSI estimates determined by CSI estimator 410.

[0155] Fig. 5 is a message flow diagram illustrating a method in accordance with example embodiments, designated generally by reference numeral 500. Method 500 may take place between a terminal device 510 (such as a user equipment) and a network node 520 (such as a gNB).

[0156] At step 530, network node 520 sends CSI-RS to terminal device 510, for measurement by terminal device 510.

[0157] At step 532, based on the received CSI-RS, terminal device 510 estimates CSI. Terminal device 510 also predicts CSI. For example, terminal device 510 may estimate CSI at the time at which a CSI-RS is received, and predict CSI at a time after a CSI-RS is received (e.g., based on at least one previous CSI-RS). The prediction may be performed by an ML based predictor.

[0158] At step 534, terminal device 510 sends a report or reports indicating predicted CSI to network node 520.

[0159] At step 536, network node 520 configures terminal device 510 to perform monitoring of CSI prediction performance. For example, by sending terminal device 510 a monitoring configuration.

[0160] At step 538, network node 520 sends CSI-RS to the terminal device.

[0161] At step 540, terminal device 510 estimates CSI from the received CSI-RS.

[0162] At step 542, terminal device 510 determines an error distribution for predicted CSI . For example, for a plurality of prediction occasions (which may correspond to occasions at which CSI-RS are received, and for which CSI is predicted), an error may be determined by comparing the prediction (e.g. based on earlier measurement performed before the prediction occasion) with the estimate (e.g. based on measurement performed during the prediction occasion). Different error metrics may be used, such as NMSE, or SGCS.

[0163] At step 544, terminal device 510 determines an error distribution for CSI predicted by a reference predictor. In some cases, the reference predictor is a ZOH predictor, taking the predicted value to be a recent estimate (such as the most recent estimate within an observation window). The reference predictor may be simpler than the predictor of step 532. The reference predictor may have fewer or simpler in-built assumptionsregarding the evolution of the channel state (for example, a ZOH predictor does not account for any evolution of the channel state, while an ML predictor may be trained to account for learned channel state behavior). Other than the predictor used, the errors determined at step 544 may be determined in a similar manner to step 542.

[0164] At step 546, terminal device 510 determines a relative performance metric by characterizing the error distribution for predicted CSI determined at step 542 with respect to the error distribution for the reference predictor determined CSI determined at step 544. For example, in some embodiments, this relative performance metric may be based on a comparison of at least one statistic of the distributions (such as the mean, mode, or median). In some embodiments, the relative performance metric is based on an area of overlap or non-overlap between the distributions of step 542 and 544. In some embodiments, the performance metric may be adjusted by further statistics (such as the variance, standard deviation, or width of one of the distributions, or a quantity derived from both distributions). In some examples, the performance metric may be bounded, modified, or adjusted based on at least one further check, such as a check that the mean of the predictor of step 542 is “better” (e.g., lower for an NMSE error statistic, or higher for an SGCS error statistic). In alternative embodiments, terminal device 510 may send the errors or distributions determined at step 542 and 544 to network node 520, for network node 520 to perform this determination of the performance metric.

[0165] At step 548, in some embodiments, a corrected performance metric is determined. For example, the performance metric may be adjusted based on a statistic of one or both distributions at this step (e.g., rather than step 546 as described above). Additionally, or alternatively, a correction factor to be applied may be determined from channel conditions, for example, using machine learning techniques. The correction factor may be determined based on measurements of the communication channel, such as measurements of the CSI-RS. In some examples, the correction factor may be determined based on one or more of: estimated CSI; an indication of the doppler spread of the communication channel; an indication of a noise level of the channel (such as a measured noise power, measured SNR, or a measured reference signal received quality, RSRQ); and an indication of an interference level of the channel (such as a determined interference power, measured signal to noise plus interference ratio, SNIR, or a measured RSRQ).

[0166] At step 550, terminal device 510, in some embodiments, sends a monitoring report to network node 520. This monitoring report may include the relative performance metric, or some indication based on the performance metric. For example, terminal device 510 may be configured to send a report containing the performance metric periodically or responsive to an indication from the network node regardless of the valueof the performance metric. In some examples, the terminal device 510 may be configured to report the performance metric based on the performance metric meeting some condition (e.g., based on a comparison of the performance metric with a configured threshold, or based on a comparison of the performance metric with a corresponding performance metric for a different predictor). In some examples, the terminal device 510 may be configured to report whether the performance metric meets a condition (e.g., periodically, responsive to an indication from the network, based on the condition being met, or based on some combination).

[0167] At step 552, network node 520 takes an action in response to the monitoring report. For example, the network node may respond by instructing terminal device 510 to use (or not use) a particular predictor, and / or by taking an ML model life cycle management decision.

[0168] Fig. 6 is a flow diagram of a method in accordance with example embodiments, designated generally by reference numeral 600. Method 600 may be carried out by a terminal device, such as a UE.

[0169] At step 610, the terminal device determines first predictions of channel state information using a first predictor. The first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions.

[0170] At step 612, the terminal device determines reference predictions of channel state information using a reference predictor. The reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions.

[0171] At step 614, the terminal device determines estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions.

[0172] At step 616, the terminal device determines first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions.

[0173] At step 618, the terminal device determines reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.By determining first and reference prediction errors, the terminal device obtains data from which the terminal device or another device may derive a relative measure of prediction performance of the first predictor.

[0174] In some example embodiments, the first predictions are determined using the first predictor based on measurements of the communication channel performed during respective observation windows that precede the respective prediction occasions. The reference predictions may be determined using the reference predictor based on measurements of the communication channel performed during respective observation windows that precede the respective prediction occasions.

[0175] In some example embodiments, the first predictor comprises an artificial intelligence, Al, and / or machine learning, ML, model.

[0176] The performance of an AI / ML model may depend on its training. Determining its performance relative to another predictor (e.g., a simpler, non-AI / ML predictor) may allow for the performance of the model to be inferred.

[0177] In some example embodiments, the reference predictor comprises a Kalman filter or a zero-order hold filter. A zero-order hold based reference predictor may reduce the processing power required to perform the reference predictions. A zero-order hold based predictor may also rely on fewer assumptions regarding the time-evolution of the channel, so the performance of the zero-order hold based predictor may be less dependent on whether those assumptions hold. A Kalman filter based predictor may rely on fewer assumptions regarding the time-evolution of the channel than an AI / ML predictor, and those assumptions may be better known than the assumptions that an AI / ML predictor may effectively learn during a training process.

[0178] In some example embodiments, the method further comprises sending, from the terminal device, information derived from the first prediction errors and the reference prediction errors to a network node. In some example embodiments, the information comprises the first prediction errors and the reference prediction errors. Sending this information to the network node may allow some of the processing burden to be borne by the network.

[0179] In some example embodiments, the method further comprises determining, by the terminal device, a prediction performance indicator by characterizing a statistical distribution of the first prediction errors with respect to a statistical distribution of the reference prediction errors.Characterizing the statistical distribution of the first prediction errors with respect to the statistical distribution of the reference prediction errors may allow the terminal device to determine an indicator or metric indicative of the relative improvement provided by the first predictor over the reference predictor.

[0180] In some example embodiments, the prediction performance indicator is determined based at least in part on a size of an area overlap and / or a size of an area of non-overlap between the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

[0181] Determining areas of overlap or non-overlap may provide a relatively computationally simple way of quantifying the relative performance of the predictors.

[0182] In some example embodiments, the determining of the prediction performance indicator is based at least in part on one or more of: a comparison of a mean of the statistical distribution of the first prediction errors and a mean of the statistical distribution of the reference prediction errors; and a comparison of a variance or standard deviation of the statistical distribution of the first prediction errors and a variance or standard deviation of the statistical distribution of the reference prediction errors.

[0183] Comparing statistics of the distributions may also provide a method of quantifying the relative performance of the predictors.

[0184] In some example embodiments, the method further comprises determining, by the terminal device, a correction factor based on at least one of the following: at least one measurement of the communication channel; a doppler spread of the communication channel; an indication of a noise level of the communication channel; and an indication of an interference level of the communication channel, and applying the correction factor to the prediction performance indicator.

[0185] A correction factor may allow the prediction performance indicator to account for broadening of the distributions that is not related to the underlying performance of the predictors.

[0186] In some example embodiments, the correction factor is determined using an AI / ML model.

[0187] For example, the correction factor may be determined using a deep neural network (DNN) ML model, having as input the real and complex values of (e.g., 52) physical resource blocks (PRBs) of the communicationchannel at at least one observation time. The DNN may be trained to perform regression of the correction factor based on a data set from a radio channel generated by a model such as the 3gpp models Uma or Umi. The objective function may be to minimize the difference between the known monitoring performance derived from ground truth CSI of the model channel and the reported performance monitoring metric generated from the reference precoder.

[0188] Using an AI / ML model may allow the correction factor to be determined accounting for a number of factors.

[0189] In some example embodiments, the terminal device further comprises means for receiving from a network node an indication of a correction factor upper and / or lower limit, and the means for determining a correction factor is configured to determine the correction factor based on said limit.

[0190] Limiting the correction factor may reduce the impact that an erroneously large or small correction factor (e.g., due to the failure of an AI / ML model used in determining the correction factor) may have on the results.

[0191] In some example embodiments, the information (derived from the prediction errors and sent to the network node) comprises the determined prediction performance indicator.

[0192] Determining the prediction performance indicator at the terminal device and sending this to the network may reduce signaling overhead.

[0193] In some example embodiments, the method further comprises: comparing, by the terminal device, the prediction performance indicator to a threshold value; and sending, from the terminal device, an indication of the result of the comparison to a network node.

[0194] Indicating the result of the comparison may convey useful information with a low signaling overhead.

[0195] In some example embodiments, the method further comprises: comparing, by the terminal device, the predictor performance indicator to a threshold value; and determining, at the terminal device, whether to use the first predictor or a different predictor based on the result of the comparison.

[0196] In some example embodiments, the method further comprises receiving, at the terminal device, an indication of the threshold value from the network.In some example embodiments, the method further comprises receiving, at the terminal device, from a network node, an indication of a number of first prediction errors and reference prediction errors from which to derive the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

[0197] In some example embodiments, the method further comprises: determining, by the terminal device, second predictions of channel state information using a second predictor, wherein the second predictions correspond to predictions of channel state information of a communication channel at the respective prediction occasions; and determining, by the terminal device, second prediction errors based on respective comparisons of the estimates of the channel state information and the second predictions of the channel state information corresponding to the respective prediction occasions.

[0198] Determining prediction errors for a second predictor may allow the terminal device to compare multiple candidate predictors to a reference predictor, and / or the prediction errors may be sent to a network node so that the network may compare multiple candidate predictors to a reference predictor.

[0199] In some example embodiments, the method further comprises determining, by the terminal device, a second prediction performance indicator by characterizing a statistical distribution of the second prediction errors with respect to a statistical distribution of the reference prediction errors; and comparing the first and second prediction performance indicators.

[0200] Comparing prediction performance indicators may allow a terminal device to select a predictor of the first and second predictors, or determine whether to switch, or indicate to the network which of the predictors has better performance.

[0201] In some example embodiments, the method further comprises selecting, by the terminal device, a predictor of the first and second predictors based on the comparison of the first and second prediction performance indicators.

[0202] In some example embodiments, the method further comprises sending, by the terminal device, an indication of the selected predictor to the network node.

[0203] In some example embodiments, the method further comprises performing, by the terminal device, data transmission or reception based at least in part on predictions of the selected predictor.Controlling terminal device behavior, such as beamforming during data transmission or reception, based on the selected predictor, may allow the terminal device to use a better performing predictor to improve performance during communication.

[0204] In some example embodiments, the method further comprises selecting the reference predictor.

[0205] Selecting the reference predictor may allow the terminal device to select a less resource intensive reference predictor (e.g. a ZOH based predictor), e.g., based on energy or processing power demands.

[0206] Fig. 7 is a flow diagram of a method in accordance with example embodiments, designated generally by reference numeral 700. Method 700 may be carried out by a network node, such as a gNB.

[0207] At step 710, the network node sends, to a terminal device, a configuration configuring the terminal device to perform method 600.

[0208] In some example embodiments, the method further comprises receiving, at the network node, from the terminal device, information derived from the first prediction errors and the reference prediction errors.

[0209] In some example embodiments, the information comprises the first prediction errors and the reference prediction errors.

[0210] This may allow the network node to determine a relative performance metric for the first predictor. Performing this determination at the network may increase the processing resources available to perform this determination, but increase signaling overhead compared to sending a prediction performance indicator.

[0211] In some example embodiments, the method further comprises determining, by the network node, a prediction performance indicator by characterizing a statistical distribution of the first prediction errors with respect to a statistical distribution of the reference prediction errors.

[0212] In some example embodiments, the information comprises a prediction performance indicator.

[0213] Receiving the prediction performance indicator at the network may reduce signaling overhead compared to receiving the prediction errors.In some example embodiments, the method further comprises comparing, by the network node, the prediction performance indicator to a threshold value.

[0214] In some example embodiments, the method further comprises receiving, at the network node, from the terminal device, an indication of the result of a comparison between the prediction performance indicator and a threshold value.

[0215] Receiving an indication of the result of a comparison at the network may reduce signaling overhead compared receiving the prediction performance errors.

[0216] In some example embodiments, the method further comprises sending, from the network node, to the terminal device, an indication of the threshold value.

[0217] This may allow the network node to tailor the threshold value to specific communication requirements.

[0218] In some example embodiments, the method further comprises sending, from the network node, to the terminal device, an indication of a number of first prediction errors and reference prediction errors from which to derive the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

[0219] This may allow the network node to tailor the sample size of the statistical distributions, and indirectly the amount of memory and processing resources required for the determination of the prediction performance indicator.

[0220] In some example embodiments the method further comprises sending, to the terminal device, an indication of a correction factor upper and / or lower limit.

[0221] This may allow the network node to tailor an upper and / or lower limit for the correction factor (e.g., which may allow the network to account for changing thresholds values for the prediction performance indicator).

[0222] For completeness, FIG. 8 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 1800. The processing system 1800 may, for example, be comprised by the device referred to in the claims below.The processing system 1800 may have a processor 1802, a memory 1804 closely coupled to the processor and comprised of a Random Access Memory (RAM) 1814 and a Read Only Memory (ROM) 1812, and, optionally, a user input 1810 and a display 1818. The processing system 1800 may comprise one or more network / apparatus interfaces 1808 for connection to a network / apparatus, e.g., a modem which may be wired or wireless. The network / apparatus interface 1808 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.

[0223] The processor 1802 is connected to each of the other components in order to control operation thereof.

[0224] The memory 1804 may comprise a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD). The ROM 1812 of the memory 1804 stores, amongst other things, an operating system 1815 and may store software applications 1816. The RAM 1814 of the memory 1804 is used by the processor 1802 for the temporary storage of data. The operating system 1815 may contain code which, when executed by the processor implements aspects of the methods 500, 600, and 700 described above. 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.

[0225] The processor 1802 may take any suitable form. For instance, it may be a microcontroller, a plurality of microcontrollers, a processor, or a plurality of processors.

[0226] The processing system 1800 may be a standalone computer, a server, a console, or a network thereof. The processing system 1800 and needed structural parts may be all inside device / apparatus such as loT device / apparatus i.e., embedded to very small size.

[0227] In some example embodiments, the processing system 1800 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 1800 may be in communication with the remote server device / apparatus in order to utilize the software application stored there.

[0228] FIG. 9 shows a tangible media, in the form of a removable memory unit 1910, storing computer-readable code which when run by a computer may perform methods according to example embodiments describedabove. The removable memory unit 1910 may be a memory stick, e.g., a Universal Serial Bus (USB) memory stick, having internal memory 1930 storing the computer-readable code. The internal memory 1930 may be accessed by a computer system via a connector 1920. 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.

[0229] Embodiments of the present invention may be implemented in digital electronic circuitry, software, hardware, application logic ora 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.

[0230] As used in this application, the term ‘circuitry’ or “circuit” may refer to one or more, or all, of the following: (a) hardware-only circuit implementations, such as implementations in analog digital circuitry, and / or quantum circuitry and (b) combinations of hardware circuit(s) and software (and / or firmware), such as (as applicable): (i) a combination of processor(s) or (ii) any or all portions of hardware processor(s) (including digital and / or quantum processor(s)), with software, and memory(ies) that work together to cause an apparatus, such as a mobile device, computing device, or server, to perform various functions, and (c) any or all portions of hardware circuit(s), such as a microprocessor(s), processor(s) and / or quantum processor(s), that require software (e.g., firmware) for operation, but the software is not necessarily present when it is not needed for operation. This definition of ‘circuitry’ applies to all uses of this term in this application. As a further example, as used in this application, the term ‘circuitry’ would also cover an implementation of merely a processor (or multiple processors) or a portion of a processor and its (or their) accompanying software and / or firmware. The term ‘circuitry’ would also cover, for example and if applicable to the particular element, a baseband integrated circuit or processor integrated circuit for a mobile phone or a similar integrated circuit in a server, a cellular network device, or another network device.

[0231] 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 specific integrated 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.

[0232] 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 above-described functions may be optional or may be combined. Similarly, it will also be appreciated that the flow and signalling diagrams of Figures 5 - 7 are examples only and that various operations depicted therein may be omitted, reordered and / or combined.

[0233] 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.

[0234] 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.

[0235] 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.

[0236] 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.

[0237] List of abbreviations

[0238] 5GC 5G Core

[0239] Al Artificial Intelligence

[0240] BVDM Building Vector Data MatrixNWDAF Network Data Analytics Function CN Core Network

[0241] CSI Channel State Information

[0242] CIR Channel Impulse Response CV Computer Vision

[0243] DCI Downlink Control Indicator

[0244] DL Downlink

[0245] DNN Deep Neural Networks

[0246] GAN Generative Adversarial Networks gNB 5G NR Node B

[0247] KPI Key Performance Indicator LSTM Long Short Term Memory MAC Medium Access Control

[0248] MAC CE MAC Control Element

[0249] MCS Modulation and Coding Scheme ML Machine Learning

[0250] MPC Multi Path Component

[0251] MSE Mean Squared Error

[0252] MSO ML-splitorchestrator

[0253] NN Neural Network

[0254] PRB Physical Resource Block

[0255] QoS Quality of Service

[0256] RAN Radio Access Network

[0257] RE Resource Element

[0258] RT Ray Tracing

[0259] RF Radio Frequency

[0260] RRC Radio Resource Control SDAP Service Data Adaptation Protocol SNR Signal to Noise Ratio

[0261] TA Tracking Area

[0262] TTI Transmission Time Interval

Claims

39ClaimsWhat is claimed is:

1. A terminal device comprising:means for determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions;means for determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions;means for determining estimates of channel state information of the communication channel at the respective prediction occasions based at least in parton respective measurements of the communication channel at the respective prediction occasions;means for determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; andmeans for determining reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

2. The terminal device of claim 1 , wherein:the first predictions are determined using the first predictor based on measurements of the communication channel performed during respective observation windows that precede the respective prediction occasions; andthe reference predictions are determined using the reference predictor based on measurements of the communication channel performed during respective observation windows that precede the respective prediction occasions.

3. The terminal device of claim 1 or claim 2, wherein the first predictor comprises an artificial intelligence, Al, and / or machine learning, ML, model.

4. The terminal device of any preceding claim, wherein the reference predictor comprises a Kalman filter or a zero-order hold filter.

405. The terminal device of any preceding claim, further comprising means for sending information derived from the first prediction errors and the reference prediction errors to a network node.

6. The terminal device of claim 5, wherein the information comprises the first prediction errors and the reference prediction errors.

7. The terminal device of any preceding claim, further comprising:means for determining a prediction performance indicator by characterising a statistical distribution of the first prediction errors with respect to a statistical distribution of the reference prediction errors.

8. The terminal device of claim 7, wherein the means for determining the prediction performance indicator is configured to determine the prediction performance indicator based at least in part on a size of an area overlap and / or a size of an area of non-overlap between the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

9. The terminal device of claim 7 or claim 8, wherein the means for determining the prediction performance indicator is configured to determine the prediction performance indicator based at least in part on one or more of:a comparison of a mean of the statistical distribution of the first prediction errors and a mean of the statistical distribution of the reference prediction errors; anda comparison of a variance or standard deviation of the statistical distribution of the first prediction errors and a variance or standard deviation of the statistical distribution of the reference prediction errors.

10. The terminal device of any of claims 7 - 9, further comprising:means for determining a correction factor based on at least one of the following:at least one measurement of the communication channel;a doppler spread of the communication channel;an indication of a noise level of the communication channel; andan indication of an interference level of the communication channel, and means for applying the correction factor to the prediction performance indicator.

11. The terminal device of claim 10, wherein the means for determining the correction factor is configured to determine the correction factor using an AI / ML model.4112. The terminal device of claim 10 or claim 11 , further comprising means for receiving from a network node an indication of a correction factor upper and / or lower limit, and wherein the means for determining a correction factor is configured to determine the correction factor based on said limit.

13. The terminal device of claim 5 and any of claims 7 - 12, wherein the information comprises the prediction performance indicator.

14. The terminal device of any of claims 7 - 13, further comprising:means for comparing the prediction performance indicator to a threshold value; andmeans for sending an indication of a result of the comparison to a network node.

15. The terminal device of any of claims 7 - 13, further comprising:means for comparing the predictor performance indicator to a threshold value; andmeans for determining whether to use the first predictor or a different predictor based on a result of the comparison.

16. The terminal device of claim 14 or claim 15, further comprising means for receiving an indication of the threshold value from the network.

17. The terminal device of claims 5 - 16, further comprising means for receiving, from a network node, an indication of a number of first prediction errors and reference prediction errors from which to derive the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

18. The terminal device of any preceding claim, further comprising:means for determining second predictions of channel state information using a second predictor, wherein the second predictions correspond to predictions of channel state information of a communication channel at the respective prediction occasions; andmeans for determining second prediction errors based on respective comparisons of the estimates of the channel state information and the second predictions of the channel state information corresponding to the respective prediction occasions.

19. The terminal device of claim 18 when dependent on claim 7, further comprising:means for determining a second prediction performance indicator by characterising a statisticaldistribution of the second prediction errors with respect to a statistical distribution of the reference prediction errors; andmeans for comparing the first and second prediction performance indicators.

20. The terminal device of any preceding claim, further comprising means for selecting the reference predictor.

21. A network node comprising:means for sending, to a terminal device, a configuration configuring the terminal device to perform a method comprising:determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions;determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions;determining estimates of channel state information at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions;determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; anddetermining reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

22. The network node of claim 21 , further comprising:means for receiving, from the terminal device, information derived from the first prediction errors and the reference prediction errors.

23. The network node of claim 22, wherein the information comprises the first prediction errors and the reference prediction errors.

24. The network node of claim 23, further comprising:means for determining a prediction performance indicator by characterising a statistical distribution of the first prediction errors with respect to a statistical distribution of the reference prediction errors.

25. The network node of claim 22, wherein the information comprises a prediction performance indicator.

26. The network node of claim 24 or 25, further comprising means for comparing the prediction performance indicator to a threshold value.

27. The network node of claim 21 , further comprising means for receiving, from the terminal device, an indication of the result of a comparison between the prediction performance indicator and a threshold value.

28. The network node of claim 27, further comprising means for sending, to the terminal device, an indication of the threshold value.

29. The network node of any of claims 21 - 28 further comprising means for sending, to the terminal device, an indication of a number of first prediction errors and reference prediction errors from which to derive the statistical distribution of the first prediction errors and the statistical distribution of the reference prediction errors.

30. The network node of any of claims 21 - 29, further comprising means for sending, to the terminal device, an indication of a correction factor upper and / or lower limit.

31. A method comprising:determining, by a terminal device, first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions;determining, by the terminal device, reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions;determining, by the terminal device, estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions;determining, by the terminal device, first prediction errors based on respective comparisons of the44estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; anddetermining, by the terminal device, reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

32. A method comprising:sending, to a terminal device, a configuration configuring the terminal device to perform a method comprising:determining first predictions of channel state information using a first predictor, wherein the first predictions correspond to predictions of channel state information of a communication channel at respective prediction occasions;determining reference predictions of channel state information using a reference predictor, wherein the reference predictions correspond to predictions of channel state information of the communication channel at the respective prediction occasions;determining estimates of channel state information of the communication channel at the respective prediction occasions based at least in part on respective measurements of the communication channel at the respective prediction occasions;determining first prediction errors based on respective comparisons of the estimates of the channel state information and the first predictions of the channel state information corresponding to the respective prediction occasions; anddetermining reference prediction errors based on respective comparisons of the estimates of the channel state information and the reference predictions of the channel state information corresponding to the respective prediction occasions.

Images