WO2025103587A1

WO2025103587A1 - First node, second node, and methods performed thereby, for handling estimations of radio signals

Info

Publication number: WO2025103587A1
Application number: PCT/EP2023/081930
Authority: WO
Inventors: Niklas WERNERSSON; Karl Werner; Rakesh Ranjan; Adrian GARCIA RODRIGUEZ
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 2023-11-15
Filing date: 2023-11-15
Publication date: 2025-05-22
Anticipated expiration: 2026-05-15

Abstract

A method performed by a first node (111). The first node (111) obtains (301), a) a plurality of estimations of noise and/or first radio signals in a first set of time-frequency resources; the signals lacking corresponding transmitted scheduling information relating to the first cell (121), and b) a plurality of estimations of second radio signals in a second set of time-frequency resources; the second radio signals being generated due to transmitted scheduling information relating to the first cell (121). The first node (111) combines (303) the pluralities to generate pairs. Each pair comprises: a) a feature comprising a subset of the plurality of estimations of the second radio signals combined with one or more estimations of noise and/or the first radio signals, and b) a respective target of the feature. The first node (111) then initiates (304) training, using machine learning, a model using, for the obtained pairs, the feature as input and the respective target as ground truth.

Description

FIRST NODE, SECOND NODE, AND METHODS PERFORMED THEREBY, FOR HANDLING ESTIMATIONS OF RADIO SIGNALS

TECHNICAL FIELD

The present disclosure relates generally to a first node and methods performed thereby for handling estimations of radio signals. The present disclosure further relates generally to a second node and methods performed thereby, for handling estimations of radio signals. The present disclosure also relates generally to computer programs and computer-readable storage mediums, having stored thereon the computer programs to carry out these methods.

BACKGROUND

Computer systems in a communications network or communications system may comprise one or more nodes. A node may comprise one or more processors which, together with computer program code may perform different functions and actions, a memory, a receiving port, and a sending port. A node may be, for example, a server. Nodes may perform their functions entirely on the cloud.

Computer systems may be comprised in a telecommunications network. The telecommunications network, communications network, or wireless communications network, may cover a geographical area which may be divided into cell areas, each cell area being served by a type of node, a network node in the Radio Access Network (RAN), radio network node or Transmission Point (TP), for example, an access node such as a Base Station (BS), e.g., a Radio Base Station (RBS), which sometimes may be referred to as e.g., gNB, evolved Node B (“eNB”), “eNodeB”, “NodeB”, “B node”, or Base Transceiver Station (BTS), depending on the technology and terminology used. The base stations may be of different classes such as e.g., Wide Area Base Stations, Medium Range Base Stations, Local Area Base Stations and Home Base Stations, based on transmission power and thereby also cell size. A cell may be understood to be the geographical area where radio coverage may be provided by the base station at a base station site. One base station, situated on the base station site, may serve one or several cells. Further, each base station may support one or several communication technologies. The telecommunications network may also comprise network nodes which may serve receiving nodes, such as user equipments or devices, with serving beams.

A radio access technology (RAT) may be understood as the connection method for a radio communications network. The RAT used may depend on the type of network nodes or devices involved in the communication.

The standardization organization Third Generation Partnership Project (3GPP) is currently in the process of specifying a New Radio Interface called Next Generation Radio or New Radio (NR), as well as a Fifth Generation (5G) Packet Core Network, which may be referred to as 5G Core Network (5GC). The advantages of 5G NR may include higher bandwidth, more resources, low latency and network slicing. 5G may provide services to various applications, such as enhanced Mobile Broad Band (eMBB), machine to Machine type communication (mMTC), Ultra Reliable Low Latency Communication (URLLC), etc.

5G may be understood to bring in sizeable flexibility with technological advancements along with innovations of cloud and Al. This may be understood to bring a whole new set of opportunities in the enterprise segment.

The wireless receiver

In a wireless communication system, radio signals may be understood to be transmitted by one communicating party via a transmitter and received by the other communicating party via a receiver. Figure 1 is a schematic diagram illustrating a transmitter/receiver in an Orthogonal Frequency Division Multiple Access (OFDM) system. In a wireless communication system such as the one illustrated in Figure 1 , the receiver may execute the following functions for estimating a channel 10 and/or receiving data.

One function may be understood to be channel estimation 11. Channel estimation may be understood as a process whereby a response of a wireless channel may be determined in order to characterize any transformation that may have been suffered by a transmitted signal prior to being received. Estimation of the radio channels may be understood to be needed to demodulate uplink transmissions, for example to coherently combine signals from multiple antennas, and correct the phase and amplitude in a frequency selective way. In NR, and Long- Term Evolution (LTE), the channel 10 may be estimated using uplink or downlink demodulation reference signals (DMRS) 12. These may be understood to be signals known to the receiver that may be interleaved in the physical resource together with data, depicted in Figure 1 in the uplink as Physical Uplink Shared Channel (PUSCH) Data 13. In Figure 1 , the reference signal 12 interleaved with the PUSCH data 13 may be transmitted by an OFDM transmitter (OFDMtx) 14. The uplink channel 10 may be understood to also be estimated via uplink sounding reference signals (SRSs), e.g., for downlink precoding computation purposes, and downlink channel state information reference signals (CSI-RSs), e.g., for downlink channel estimation in frequency division multiplexing (FDD) systems.

The transmitted signal may be received by a receiver (Rx), where it may first undergo signal processing at the Rx 15, which in this OFDM system comprises OFDM demodulation.

Another function may be understood to be equalization 16. When the receiver may attempt to decode data, it may typically rely on the estimated wireless channel to equalize the received signals, that is, to remove the impact of the wireless channel 10, noise, and interference 17 and estimate the transmitted data symbols. Yet another function may be understood to be soft demapping 18. After the equalizer, the receiver may typically compute the soft bits by calculating the probability that the transmitted bits were 0 or 1 , conditioned on the estimates of the transmitted data symbols after equalization 16.

A further function may be understood to be decoding 19. The output of the soft demapper may be used by the decoder to perform hard decisions on whether the transmitted bits were 0 or 1 .

In the schematic of Figure 1 , it may be noted that the received signal may consist of a linear combination of the transmitted signal, filtered through the channel 10, and the noise/interference 17. It may be noted also that contributions from noise/interference 17 and channel part may add, approximately, linearly at cut 2. Another relevant aspect may be understood to be that interference/noise 17 may be understood to vary in power, and that it may be understood to be rather independent from how the channel 10, to served User Equipment (UE), may vary. Cut 2.1 may be understood to indicate the resources, e.g., resource elements in the case of the figure, including reference signals for channel estimation purposes. Cut 2.2 may be understood to indicate the resources, e.g., resource elements in the case of the figure, including data.

Artificial Intelligence (Al) in telecom

Al is considered as an enabler of enhancements in the future generation network, e.g., 6G, and may be regarded as key leverage to transform the whole design philosophy to a new level of adaptivity to customize radio systems for diverse radio environments. This may be understood to be especially important given the growing complexity in RANs with each new generation. The learning capabilities of Al may be understood to create advantageous policy or strategies directly based on data, instead of human logics and symbolic modelling and analysis.

Machine learning (ML) may be understood as the study of computer algorithms that may improve automatically through experience. It is seen as a part of Al. ML algorithms may build a model based on sample data, known as "training data", in order to make predictions or decisions without being explicitly programmed to do so. ML algorithms may be used in a wide variety of applications, such as email filtering and computer vision, where it may be difficult or unfeasible to develop conventional algorithms to perform the needed tasks.

AI/ML enabled methods may be understood to employ data-driven learning approaches where the models may learn the underlying data distribution and relationships between the inputs and outputs without the need for understanding the inherent complex processes. AI/ML enabled methods may be understood to mainly rely on statistical techniques.

On the contrary, legacy methods may be understood to be model-driven, where a method may be derived based on a simplified model of the underlying problem. In the area of channel estimation one example of this may be the legacy algorithms based on least squares or linear minimum mean squared error (L-MMSE).

In ML, there may be basically three types of ML Algorithms: Supervised Learning, Unsupervised Learning, and Reinforcement Learning (RL).

Supervised Learning algorithms may comprise a target/outcome variable, or dependent variable, which may have to be predicted from a given set of predictors, that is, independent variables. Using this set of variables, a function may be generated that may map inputs to desired outputs. The training process may continue until the model may achieve a desired level of accuracy on the training data. Once an ML model may have been trained, an inference process may begin, whereby new data may be run through the ML model to calculate an output. Examples of Supervised Learning may be Regression, Decision Tree, Random Forest, KNN, Logistic Regression etc.

In Unsupervised Learning algorithms, there may be no target or outcome variable to predict/estimate. It may be used for clustering a population into different groups, which may be widely used for segmenting customers in different groups for specific intervention. Examples of Unsupervised Learning may be K-means, mean-shift clustering, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Expectation-Maximization (EM) Clustering using Gaussian Mixture Models (GMM), Agglomerative Hierarchical Clustering, etc....

Cluster analysis or clustering may be understood as an ML technique which may comprise grouping a set of objects in such a way that objects in the same group, which may be called a cluster, may be understood to be more similar, in some sense, to each other than to those in other groups, that is, other clusters. It may be understood as a main task of exploratory data mining, and a common technique for statistical data analysis, used in many fields, including pattern recognition, image analysis, information retrieval, bioinformatics, data compression, computer graphics and ML.

Using an RL algorithm, a machine may be trained to make specific decisions. It may be understood to work as follows: the machine may be exposed to an environment where it may train itself continually using trial and error. This machine may learn from past experience and may try to capture the best possible knowledge to make accurate decisions. An example of RL may be a Markov Decision Process (MDP). The training using RL may comprise generating an ML model. To train such an ML model, an agent, given a state of the environment, may take an action in this environment and receive a reward. The action may result in a new state of the environment. This process may be repeated in a loop. Over time, the agent may learn to take actions that may result in larger immediate and future rewards, meaning that it may be understood to be in the best interest of the agent not to take the action that may only lead to the highest reward in the next state, but the action that may cumulatively lead to the highest reward in the next state and in a future number of states. The agent may comprise a neural network which may input the state and may produce an action. There may be several ML algorithms that may be used for training the network of the agent, e.g., policy-learning based, such as actor-critic approaches, or value-based learning, such as deep-q networks.

An approach based on data-driven learning may be understood to be a promising application of machine learning for the physical layer at radio system, such as a path to design transceivers adaptive to radio environments, possible performance gains over a general modelling or an inaccurate modelling-based system, reducing product design cycles with more general modelling, etc.

At the receiver, one or more of the functions mentioned above, that is, channel estimation, equalization, soft demapping and decoding, may be executed by AI/ML models separately. That is, a single ML model may replace a single functionality such as channel estimation, or jointly. That is, a single ML model may replace several functionalities to perform a joint channel estimation, equalization, and soft demapping. It may be noted that a single AI/ML model solely performing channel estimation or a single AI/ML model performing all four functionalities described above, or several AI/ML models, e.g., a receiver may comprise a first AI/ML model jointly performing a joint channel estimation, equalization, and soft demapping, and a second AI/ML model performing data decoding.

Meanwhile, at the current stage, the data-driven approaches may possibly incur challenges, such as generalizability of the scheme used, especially for offline training, and its training efficiency, and feedback loop availability and quality for online training & efficiency.

SUMMARY

As part of the development of embodiments herein, one or more problems with the existing technology will first be identified and discussed.

As shown in simulations, AI/ML-based functionalities may be understood to be capable of providing better performance than legacy methods. However, the ML models may be understood to need to be trained with relevant channel data, including those having high interference/noise. Otherwise, the accuracy of the models may be reduced in deployment.

One persistent challenge with ML, especially deeper ML models, is the need of substantial amounts of high-quality data. The need for data increases rapidly with the size of the model, and access to data may often be limited. The collection of high-quality data poses challenges.

In some cases, it may be expected that the data may have to be captured in the same sector the trained ML model may be expected to be deployed in. It may be beneficial if the data is representative of the actual channel realizations that may be seen in the cell/sector. To train AI/ML models at the receiver performing one or more functionalities, the AI/ML models may typically need to be trained with many pairs of data samples. Each pair may comprise the input/s to the AI/ML model, and the targeted output/s of the AI/ML, that is, the labels that may be used as ground truths. Examples of inputs may include at least the received reference signals for channel estimation in the case of an AI/ML model for channel estimation, or at least the received reference signals for channel estimation, if present, and the received data signals in the case of an AI/ML model for joint channel estimation, equalization, and soft demapping. Examples of labels may include the actual channel response, without noise or interference, in the case of an AI/ML model for channel estimation, or the transmitted data bits, that is, the correctly decoded data bits, in the case of an AI/ML model for joint channel estimation, equalization, and soft demapping.

In AI/ML models, a feature may be understood to refer to as one or more variables provided as input. While the features may be typically readily obtained from the received signal after going through the receiver functionalities before the input to the AI/ML model, obtaining the target may be understood to require some means of, e.g., removing noise/interference from the received reference signals in the case of AI/ML channel estimation, or guaranteeing a correct decoding of the transmitted data bits in the case of an AI/ML model for joint channel estimation, equalization, and soft demapping. This may prove to be rather challenging, especially when interference/noise is strong in relation to the signal.

Embodiments herein may address the problems of the existing methods just described.

According to a first aspect of embodiments herein, the object is achieved by a computer- implemented method, performed by a first node. The method is for handling estimations of radio signals. The first node operates in a computer system. The first node obtains, directly or indirectly, for a plurality of radio time-frequency resources configured for use for communications between one or more devices operating, or having operated, in a communications system, and one or more cells in the communications system the following two pluralities of estimations. The first node obtains a plurality of estimations of noise and/or first radio signals in a first set of time-frequency resources in a first set of time-frequency resources. The first set of time-frequency resources are configured for use in a first cell of the one or more cells. The first radio signals lack corresponding transmitted scheduling information relating to the first cell. The first node also obtains, a plurality of estimations of second radio signals in a second set of time-frequency resources configured for use in the first cell. The second radio signals are generated due to transmitted scheduling information relating to the first cell. The first node combines the plurality of estimations of noise and/or the first radio signals with one or more estimations of the plurality of estimations of the second radio signals to generate a plurality of pairs. Each pair in the plurality of pairs comprises a feature. The feature comprises a subset of the plurality of estimations of the second radio signals combined with one or more estimations of noise and/or the first radio signals. The pair also comprises a respective target of the feature. The respective target of the feature is one of: a) obtained by applying the one or more functions to one of the estimations in said subset, and b) one of the estimations in the subset. The first node also initiates training, using machine learning, of a mathematical model of estimations of radio signals in radio time-frequency resources. The training is performed using, for the obtained plurality of pairs, the feature as input and the respective target as ground truth.

According to a second aspect of embodiments herein, the object is achieved by a computer-implemented method, performed by the second node. The method is for handling estimations of radio signals. The second node operates in the computer system. The second node obtains, directly or indirectly, for a fourth set of time frequency resources configured for use for communications between a second device operating, or having operated, in the communications system, and a second cell in the communications system one or more measurements. The second node receives the first indication from the first node operating in the computer system. The first indication indicates the mathematical model of estimations of radio signals in radio time-frequency resources, trained, using machine learning. The training has been performed using, for the obtained plurality of pairs described in relation to the first node, the feature as input and the respective target as ground truth. The second node then uses the trained mathematical model to estimate a further radio signal in the fourth set of time frequency resources.

According to a third aspect of embodiments herein, the object is achieved by the first node. The first node may be understood to be for handling the estimations of radio signals. The first node is configured to operate in the computer system. The first node is configured to obtain, directly or indirectly, for the plurality of radio time-frequency resources configured for use for communications between the one or more devices being configured to be operating, or configured to have operated, in the communications system, and the one or more cells in the communications system the following. The first node is configured to obtain the plurality of estimations of noise and/or first radio signals in the first set of time-frequency resources configured for use in the first cell of the one or more cells. The first radio signals are configured to lack corresponding transmitted scheduling information relating to the first cell. The first node is also configured to obtain the plurality of estimations of the second radio signals in the second set of time-frequency resources configured for use in the first cell 121. The second radio signals are configured to be generated due to transmitted scheduling information relating to the first cell. The first node is also configured to combine the plurality of estimations of noise and/or the first radio signals with the plurality estimations of the plurality of estimations of the second radio signals to generate the plurality of pairs. Each pair in the plurality of pairs is configured to comprise: i) the feature configured to comprise the subset of the plurality of estimations of the second radio signals combined with the one or more estimations of noise and/or the first radio signals, and ii) the respective target of the feature. The respective target of the feature is configured to be one of: a) obtained by applying one or more functions to one of the estimations in the subset, and ii) the one of the estimations in the subset. The first node is further configured to initiate training, using machine learning, of the mathematical model of estimations of radio signals in radio time-frequency resources. The training is configured to be performed using, for the plurality of pairs configured to be obtained, the feature as input and the respective target as ground truth.

According to a fourth aspect of embodiments herein, the object is achieved by the second node. The second node may be understood to be for handling estimations of radio signals. The second node is configured to operate in the communications network. The second node is configured to obtain, directly or indirectly, for the fourth set of time frequency resources configured for use for communications between the second device configured to operate, or configured to have operated, in the communications system, and the second cell in the communications system the one or more measurements. The second node is also configured to receive the first indication from the first node configured to operate in the computer system. The first indication is configured to indicate the mathematical model of estimations of radio signals in radio time-frequency resources described in relation to the first node, configured to be trained, using machine learning. The training is configured to have been performed using, for the plurality of pairs configured to be obtained as described in relation to the first node, the feature as input and the respective target as ground truth. The second node is further configured to use the trained mathematical model to estimate the further radio signal in the fourth set of time frequency resources.

According to a fifth aspect of embodiments herein, the object is achieved by a computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the first node.

According to a sixth aspect of embodiments herein, the object is achieved by a computer-readable storage medium, having stored thereon the computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the first node.

According to a seventh aspect of embodiments herein, the object is achieved by a computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the second node.

According to an eighth aspect of embodiments herein, the object is achieved by a computer-readable storage medium, having stored thereon the computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the second node.

By obtaining the plurality of estimations of noise and/or the first radio signals and the one or more estimations of the plurality of estimations of the second radio signals, the first node may then be enabled to combine the estimations.

By combining the plurality of estimations of noise and/or the first radio signals with one or more estimations of the plurality of estimations of the second radio signals to generate the plurality of pairs of features and their respective targets, the first node may enable to then train, using machine learning, a mathematical model with the combined pairs, so that the mathematical model may learn to identify the target from an inputted feature. The first node may enable that training data for a ML-based mathematical model may be constructed so that important and challenging low SINR cases may be guaranteed to be sufficiently represented. The channel realizations may be well estimated because they may be captured in high SINR, where channel estimation may be easier. The high interference/noise cases may also be captured. The training data set may be enabled to be larger because any combination of noise/interference and channel data may be used to form a training sample, so that data augmentation may be facilitated. All these advantages may be understood to lead to a better ML solution.

By initiating the training of the mathematical model, the first node may then enable that, once the mathematical model may have been trained, the mathematical model may be used to, e.g., perform the function for which the mathematical model may have been trained, e.g., channel estimation, or channel estimation, equalization, and soft demapping, on any input received signal, with a high level of accuracy. As will be shown with an example in Figure 5, ML based channel estimation has been shown to outperform the legacy AIC based channel estimator.

BRIEF DESCRIPTION OF THE DRAWINGS

Examples of embodiments herein are described in more detail with reference to the accompanying drawings, according to the following description.

Figure 1 is a schematic diagram illustrating a transmitter/receiver in an OFDM system, according to existing methods.

Figure 2 is a schematic diagram illustrating two non-limiting examples, in panels a) and b), of a computer system, according to embodiments herein.

Figure 3 is a flowchart depicting a method in a first node, according to embodiments herein. Figure 4 is a schematic diagram depicting aspects of a method performed by the first node, according to embodiments herein. Figure 5 is a schematic diagram depicting experimental results obtained using a method performed by the first node, according to embodiments herein.

Figure 6 is a schematic diagram depicting further aspects of a method performed by the first node, according to embodiments herein.

Figure 7 is a flowchart depicting a method in a second node, according to embodiments herein.

Figure 8 is a schematic block diagram illustrating an embodiment of a first node, according to embodiments herein.

Figure 9 is a schematic block diagram illustrating an embodiment of a second node, according to embodiments herein.

DETAILED DESCRIPTION

Certain aspects of the present disclosure and their embodiments address the challenges identified in the Background and Summary sections with the existing methods and provide solutions to the challenges discussed.

Embodiments herein may be understood to relate to data generation from network realizations for training of ML models.

Some of the embodiments contemplated will now be described more fully hereinafter with reference to the accompanying drawings, in which examples are shown. In this section, the embodiments herein will be illustrated in more detail by a number of exemplary embodiments. Other embodiments, however, are contained within the scope of the subject matter disclosed herein. The disclosed subject matter should not be construed as limited to only the embodiments set forth herein; rather, these embodiments are provided by way of example to convey the scope of the subject matter to those skilled in the art. It should be noted that the exemplary embodiments herein are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments.

Several embodiments and examples are comprised herein. It should be noted that the embodiments and/or examples herein are not mutually exclusive. Components from one embodiment or example may be tacitly assumed to be present in another embodiment or example and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments and/or examples.

Figure 2 depicts two non-limiting examples, in panels “a” and “b”, respectively, of a computer system 100, in which embodiments herein may be implemented. In some example implementations, such as that depicted in the non-limiting examples of Figure 2 a), the computer system 100 may be a computer network. In other example implementations, such as that depicted in panel b) of Figure 2, the computer system 100 may be implemented in a communications system 101 , that is, a telecommunications system, sometimes also referred to as a cellular radio system, cellular network or wireless communications system. In some examples, the communications system 101 may comprise network nodes which may serve receiving nodes, such as wireless devices, with serving beams.

In some examples, the communications system 101 may be, for example, a communications network, such as 5G system, or Next Gen network. The communications system 101 may also, or alternatively, support other technologies, such as LTE, e.g., LTE Frequency Division Duplex (FDD), LTE Time Division Duplex (TDD), LTE Half-Duplex Frequency Division Duplex (HD-FDD), LTE operating in an unlicensed band, Wideband Code Division Multiple Access (WCDMA), Universal Terrestrial Radio Access (UTRA) TDD, Global System for Mobile communication (GSM)ZEnhanced Data Rate for GSM Evolution (EDGE) Radio Access Network (GERAN) network, Ultra-Mobile Broadband (UMB), EDGE network, network comprising of any combination of Radio Access Technologies (RATs) such as e.g., Multi-Standard Radio (MSR) base stations, multi-RAT base stations etc., any 3rd Generation Partnership Project (3GPP) cellular network, Wireless Local Area Network/s (WLAN) or WiFi network/s, Worldwide Interoperability for Microwave Access (WiMax), Internet of Things (loT), Machine Type Communication (MTC), IEEE 802.15.4-based low-power short-range networks such as IPv6 over Low-Power Wireless Personal Area Networks (6LowPAN), Zigbee, Z-Wave, Bluetooth Low Energy (BLE), or any cellular network or system.

The computer system 100 comprises nodes, whereof a first node 111 and a second node 112 are depicted in Figure 2. In some examples, not depicted in Figure 2, the first node 111 and the second node 112 may be co-located or be the same node. The computer system 100 may comprise additional nodes.

Any of the first node 111 and the second node 112 may be understood, respectively, as a first computer system or server and a second computer system or server. Any of the first node 111 and the second node 112 may be implemented as a standalone server in e.g., a host computer in the cloud 115, as depicted in the non-limiting example of Figure 2b). In some examples, any of the first node 111 and the second node 112 may be a distributed node or distributed server, such as a virtual node in the cloud 115, and may perform some of its respective functions locally, e.g., by a client manager, and some of its functions in the cloud 115, by e.g., a server manager. In other examples, any of the first node 111 and the second node 112 may perform its functions entirely on the cloud 115, or partially, in collaboration or collocated with a radio network node. Yet in other examples, any of the first node 111 and the second node 112 may also be implemented as processing resources in a server farm. In yet other examples, any of the first node 111 and the second node 112 may be comprised in a device, such as any of the one or more devices 130 described below, or on the edge. Any of the first node 111 and the second node 112 may be under the ownership or control of a service provider or may be operated by the service provider, or on behalf of the service provider.

Any of the first node 111 , and the second node 112 may have a capability to perform machine-implemented learning procedures, which may be also referred to as “machine learning” (ML).

Any of the first node 111 , and the second node 112 may have a capability to manage an artificial neural network. The artificial neural network may be understood as a machine learning framework, which may comprise a collection of connected nodes, where in each node or perceptron, there may be an elementary decision unit. Each such node may have one or more inputs and an output. The input to a node may be from the output of another node or from a data source. Each of the nodes and connections may have certain weights or parameters associated with it. In order to solve a decision task, the weights may be learnt or optimized over a data set which may be representative of the decision task. The most commonly used node may have each input separately weighted, and the sum may be passed through a non-linear function which may be known as an activation function. The nature of the connections and the node may determine the type of the neural network, for example a feedforward network, recurrent neural network etc. That any of the first node 111 and the second node 112 may have the capability to manage the artificial neural network may be understood herein as having the capability to store the training data set and the models that may result from the machine learning, to train a new model, and once the model may have been trained, to use this model for prediction. In some embodiments, such as those depicted in Figure, the system that may be used for training the model and the one used for prediction may be different. That is, the first node 111 may be different from the second node 112. In some embodiments, the first node 111 may be understood as a node having a capability to use the model, once the model may have already been trained. However, in other embodiments, the first node 111 may also have a capability to train the artificial neural network 121.

Any of the second node 112, and in some embodiments, the first node 111 , used for training the artificial neural network, may require more computational resources than the first node 111 that may use the trained model to make predictions. Therefore, any of the second node 112, and in some embodiments, the first node 111 , used for training the artificial neural network may, for example, support running python/Java with Tensorflow or Pytorch, Theano etc... Any of the first node 111 and the second node 112 may also have GPU capabilities.

In some embodiments, any of the first node 111 and the second node 112 may be a core network node, such as, e.g., a network data analytics function (NWDAF), a Service management and orchestration (SMO) node, a positioning node, a coordinating node, a Self- Optimizing/Organizing Network (SON) node, a Minimization of Drive Test (MDT) node, etc.... In 5G, for example, any of the first node 111 and the second node 112 may be located in the Operations Support Systems (OSS).

In some examples not depicted in Figure 2, any of the first node 111 and the second node 112 may device, such as any of one or more devices 130 comprised in the communications system 101 depicted in Figure 2. In the non-limiting example depicted in Figure 2, the one or more devices comprise a first device 131 and a second device 132. This may be understood to be for illustrative purposes only, and non-limiting. Any of the one or more devices 130 comprised in the communications system 101 may be a wireless communication device such as a 5G UE, or a UE, which may also be known as e.g., mobile terminal, wireless terminal and/or mobile station, a Customer Premises Equipment (CPE) a mobile telephone, cellular telephone, or laptop with wireless capability, just to mention some further examples. The one or more devices 130 comprised in the communications system 101 may be, for example, portable, pocket-storable, hand-held, computer-comprised, or a vehiclemounted mobile device, enabled to communicate voice and/or data, via the RAN, with another entity, such as a server, a laptop, a Personal Digital Assistant (PDA), or a tablet, Machine-to- Machine (M2M) device, device equipped with a wireless interface, such as a printer or a file storage device, modem, sensor, loT device, or any other radio network unit capable of communicating over a radio link in the communications system 101. The one or more devices 130 comprised in the communications system 101 may be enabled to communicate wirelessly in the communications system 101. The communication may be performed e.g., via a RAN, and possibly the one or more core networks, which may be comprised within the communications system 101.

In other examples, any of the first node 111 and the second node 112 may be a radio network node serving the one of the one or more devices 130. In the non-limiting example of Figure 2 b), the first device 131 is served by a first radio network node 141 and the second device 132 is served a second radio network node 142. Hence, the radio network node serving the one or more devices 130 may be referred to herein as radio network node 141, 142. The radio network node 141 , 142 may be, e.g., comprised in a Radio Access Network of the communications system 101. That is, the radio network node 141 , 142 may be a transmission point such as a radio base station, for example a gNB, an eNB, or any other network node with similar features capable of serving a wireless device, such as a user equipment or a machine type communication device, in the communications system 101. In typical examples, the radio network node 141 , 142 may be a base station, such as a gNB or an eNB. In other examples, the radio network node 141 , 142 may be a distributed node, such as a virtual node in the cloud 115, and may perform its functions entirely on the cloud 115, or partially, in collaboration with a radio network node.

The communications system 101 may cover a geographical area, which in some embodiments may be divided into cell areas, wherein each cell area may be served by a radio network node 141 , 142, although, one radio network node 141 , 142 may serve one or several cells. The communications system 101 comprises one or more cells 120, whereof a first cell 121 and a second cell 122 are depicted in the non-limiting example of Figure 2b. In the nonlimiting example of Figure 2, the first radio network node 141 serves the first cell 121 and the second radio network node 142 serves the second cell 122. The radio network node 141 , 142 may be of different classes, such as, e.g., macro eNodeB, home eNodeB or pico base station, based on transmission power and thereby also cell size. In some examples, the radio network node 141 , 142 may serve receiving nodes with serving beams. The radio network node 141 , 142 may be directly connected to one or more core networks.

Any of the first node 111 , the second node 112 and the radio network node 141 , 142 comprised in the communications system 101 may support one or several communication technologies, and its name may depend on the technology and terminology used.

It may be understood that the communications system 101 may comprise additional radio network nodes and/or additional devices.

Any of the one or more devices 130 and the radio network node 141 , 142 may have a receiver 150 to receive radio signals. In the non-limiting example depicted in Figure 2 a) the first device 131 has a first receiver 151 and the second device 132 has a second receiver 152. In the non-limiting example depicted in Figure 2 b) the first radio network node 141 has a third receiver 153 and the second radio network node 154 has a fourth receiver 154.

The first node 111 may be configured to communicate within the computer system 100 with the second node 112 over a first link 161 , e.g., a radio link, or a wired link. The first node 111 may be configured to communicate within the communications system 101 with the one or more devices 130 over a respective link. In the non-limiting example of Figure 2 a), the first node 111 may be configured to communicate within the communications system 101 with the first device 131 over a second link 162, e.g., a radio link, or a wired link. The first node 111 may be configured to communicate within the communications system 101 with the second device 132 over a third link 163, e.g., a radio link, or a wired link. The first node 111 may be configured to communicate within the communications system 101 with radio network node 141 , 142 over another respective link. In the non-limiting example of Figure 2 a), the first node 111 may be configured to communicate within the communications system 101 with the first radio network node 141 over a fourth link 164, e.g., a radio link. In the non-limiting example of Figure 2 a), the first node 111 may be configured to communicate within the communications system 101 with the second radio network node 142 over a fifth link 165, e.g., a radio link, or a wired link. The radio network node 141 , 142 may be configured to communicate within the communications system 101 with the one or more devices 130 over a further respective link. In the non-limiting example of Figure 2 a), the first radio network node 141 may be configured to communicate within the communications system 101 with the first device 131 over a sixth link 166, e.g., a radio link, or a wired link. In the non-limiting example of Figure 2 a), the second radio network node 142 may be configured to communicate within the communications system 101 with the second device 132 over a seventh link 167, e.g., a radio link, or a wired link.

Any of the first link 161 , the second link 162, the third link 163, the fourth link 164, the fifth link 165, the sixth link 166 and the seventh link 167 may be a direct link or may be comprised of a plurality of individual links, wherein it may go via one or more computer systems or one or more core networks in the computer system 100, which are not depicted in Figure 2, or it may go via an optional intermediate network. The intermediate network may be one of, or a combination of more than one of, a public, private or hosted network; the intermediate network, if any, may be a backbone network or the Internet; in particular, the intermediate network may comprise two or more sub-networks, which is not shown in Figure 2.

In general, the usage of “first”, “second”, “third”, “fourth”, “fifth”, “sixth” and/or “seventh”, herein may be understood to be an arbitrary way to denote different elements or entities, and may be understood to not confer a cumulative or chronological character to the nouns they modify.

Some of the embodiments contemplated herein will now be described more fully with reference to the accompanying drawings. Other embodiments, however, are contained within the scope of the subject matter disclosed herein, the disclosed subject matter should not be construed as limited to only the embodiments set forth herein; rather, these embodiments are provided by way of example to convey the scope of the subject matter to those skilled in the art.

Embodiments of a computer-implemented method, performed by the first node 111 , will now be described with reference to the flowchart depicted in Figure 3. The method is for handling estimations of radio signals. The first node 111 operates in the computer system 100.

Several embodiments are comprised herein. In some embodiments all the actions may be performed. In some embodiments, some actions may be optional. In Figure 3, optional actions are indicated with dashed lines. It should be noted that the examples herein are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description.

Action 301

Embodiments herein may be understood to relate to ultimately enabling estimation of transmitted signals by using a mathematical model trained with machine learning, that is, by using an AI/ML model. The mathematical model may have different functionality. For example, in a first group of examples, a first functionality of the mathematical model may be for joint channel estimation, equalization, and soft demapping. In a second group of examples, a second functionality of the mathematical model may be for channel estimation.

In order to enable the training of such an AI/ML model, the first node 111 may first obtain data in this Action 301 , in order to determine inputs to train the AI/ML model. Some aspects to the collection of data in this Action 301 may depend on the functionality of the mathematical model to be trained, as will be noted.

In this Action 301 , the first node 111 obtains, directly or indirectly, for a plurality of radio time-frequency resources configured for use for communications between the one or more devices 130 operating, or having operated, in the communications system 101 , and the one or more cells 120 in the communications system 101 the following two pluralities of estimations: a plurality of estimations of noise and/or first radio signals in a first set of time-frequency resources and a plurality of estimations of second radio signals in a second set of timefrequency resources configured for use in the first cell 121 , explained in further detail below.

In some examples, obtaining in this Action 301 may be understood as receiving. The receiving may be, in some examples, from the radio network node, network node or device having performed the estimations, e.g., any of the one or more devices 130 in the first cell 121 , such as the first device 131 , or the first radio network node 141 serving the first cell 121. In other examples, the receiving may be from a core network node, where the estimations may have been calculated.

In other examples, the obtaining in this Action 301 may be understood as fetching, e.g., from another node or memory in the computer system 100, where the estimations may be stored.

The terms estimation and estimate are used indistinctively herein.

Yet it other examples, the obtaining in this Action 301 may comprise performing the estimations. That is, performing the calculations, derivations or similar. In some of such examples, the first node 111 may, for example, be co-localized with the receiver 150 of the radio signals. In some embodiments, the first node 111 may be one of the one or more devices 130 or the radio network node 141 , 142 serving one of the one or more devices 130. Particularly, in some embodiments, the first node 111 may be one or the first device 131 in the first cell 131 and the first radio network node 141 serving the first cell 121.

That is, embodiments herein may be applied at the network, such as at the first radio network node 141 , e.g., acquiring data for training models involved in the reception of uplink signals, or at the device, such as at the first device 131 , e.g., acquiring data for training models involved in the reception of downlink signals.

The cell 121 , as used herein may refer to a sector related to the first cell 121 .

The radio time-frequency resources may be, for example resource elements (REs). Configured for use in the first cell 121 may be understood to mean any of allocated, assigned or reserved for use in the first cell 121 . However, as will be explained later, this does not necessarily mean that there are scheduled devices in the radio time-frequency resources, as will be explained below.

In some embodiments, any of the estimations, that is, of any of the two pluralities of estimations, may be performed using synthetic channel realizations. In such embodiments, the set of noise and/or interference realizations may be instead synthetic in the sense that it may be obtained from a digital representation of the communications network 101 , e.g., a digital twin, of the communications network 101. It may, for instance, be so that information about the propagation environment may be obtained, e.g., information about buildings, trees, roads, gNB locations and potential UE locations, and based on this, a model may be created which may describe noise and/or interference realizations in a given sector.

Estimations of noise and/or first radio signals

In this Action 301 , the first node 111 obtains the plurality of estimations of noise and/or first radio signals in a first set of time-frequency resources. The first set of time-frequency resources are configured for use in the first cell 121 of the one or more cells 120. The first set of time-frequency resources may be understood to be a set of the plurality of radio timefrequency resources configured for use for communications between the one or more devices 130 operating, or having operated, in the communications system 101 , and the one or more cells 120 in the communications system 101.

The first radio signals lack corresponding transmitted scheduling information relating to the first cell 121 . That is, in some examples, the obtaining of the plurality of estimations of noise and/or first radio signals may be performed when no user may be scheduled, that is when only noise and/or interference may be received. Accordingly, the first radio signals may be understood to not originate from scheduled devices in the first cell 121 . The first radio signals may, however, originate from devices, e.g., scheduled devices, in other cells, e.g., nearby cells, such as the second cell 122 in Figure 2b. For example, in Figure 2b, the first radio signals may originate form the second device 132. Accordingly, the obtaining of the plurality of estimations of noise and/or first radio signals in the first set of time-frequency resources may be understood as obtaining an estimation, or storing, a set of noise and/or interference realizations.

In some embodiments, the obtaining in this Action 301 of the plurality of estimations of noise and/or first radio signals may comprise at least one of the following. According to a first option, the obtaining in this Action 301 of the plurality of estimations of noise and/or first radio signals may comprise performing measurements of the noise and/or first radio signals when the first set of time-frequency resources lack corresponding transmitted scheduling information relating to the first cell 121. In some of these examples, the first node 111 may be co-localized with the receiver 150 of the noise and/or first radio signals. According to a second option, the obtaining in this Action 301 of the plurality of estimations of noise and/or first radio signals may comprise removing intended signal from respective measured signals in the first set of timefrequency resources 155. In such embodiments, the noise and/or interference realizations may be obtained by removing the received useful signal from the input of the relevant AI/ML functionality as explained above, e.g., by removing the estimated channel of the sector links from the signal measured in Cut 2.1 within Figure 4. The input of the relevant AI/ML functionality may be understood to be used to signify at which point of the receiver the signal may be introduced to the AI/ML functionality, e.g., “Cut” in the figure. This may be possible when the received useful signal at the inputs of the relevant AI/ML functionalities may be reliably estimated, e.g., when the Signal-to-lnterference Noise Ratio (SINR) may be sufficient. A useful signal may be understood as the signal intended to be received/decoded, that is, the signal intended to the receiver.

Accordingly, in some embodiments, the first radio signals may have a measured power below a first threshold. The power threshold may be a reference signal received power threshold.

According to a third option, the obtaining in this Action 301 of the plurality of estimations of noise and/or first radio signals may comprise performing the measurements between a demodulator and a demapper of the communication receiver 150 used to receive the measured signals. For example, the realizations of the noise and/or first radio signals may be measured in Cut 2.1 in Figure 4 for a case that the AI/ML-based mathematical model may have to be eventually trained for channel estimation or in Cut 2 in Figure 4 for a case that the AI/ML-based mathematical model may have to be eventually trained for joint channel estimation, equalization, and soft demapping. Figure 4 will be described later.

In some embodiments noise and/or interference samples may be obtained in a manner such that they may be sampled and/or categorized in different load levels.

By in this Action 301 obtaining the plurality of estimations of noise and/or first radio signals in a first set of time-frequency resources, the first node 111 may be enabled to obtain, e.g., estimate and/or store a set of AI/ML model inputs, relevant to links between the first cell 121 , e.g., sector and the devices served by the first cell 121 , e.g., the first device 131 , wherein the high interference/noise cases may also be captured.

Estimations of second radio signals

In this Action 301 , the first node 111 also obtains, the plurality of estimations of second radio signals in a second set of time-frequency resources configured for use in the first cell 121. The second set of time-frequency resources may be understood to be another set of the plurality of radio time-frequency resources configured for use for communications between the one or more devices 130 operating, or having operated, in the communications system 101 , and the one or more cells 120 in the communications system 101.

The second radio signals are generated due to transmitted scheduling information relating to the first cell 121 . That is, the second radio signals may be understood to originate from scheduled devices in the first cell 121. Accordingly, the obtaining of the plurality of estimations of second radio signals in the second set of time-frequency resources may be understood as obtaining an estimation, or storing, channel realizations, e.g., which may be understood as useful signals, without or with limited noise/interference, relevant to links between the first cell 121 and the devices in the first cell 131 , the first device 131.

The obtaining of the plurality of estimations of second radio signals

In some embodiments, the obtaining in this Action 301 of the plurality of estimations of the second radio signals may be based on measurements performed in one of the following. According to a first option, the obtaining in this Action 301 of the plurality of estimations of the second radio signals may be based on measurements performed between an output of a demodulator and an input to a decoder of a communication receiver 150 used to receive the measured signals. This may be understood to correspond to Cut 2.1 within Figure 6 for the case of AI/ML-based channel estimation, which will be described later. According to a second option, the obtaining in this Action 301 of the plurality of estimations of the second radio signals may be based on measurements performed at an input of a channel estimator of the communication receiver 150 used to receive the measured signals. This may be understood to correspond to “Cut 2.1” in Figure 6, which will be described later. According to a third option, the obtaining in this Action 301 of the plurality of estimations of the second radio signals may be based on measurements performed at an output of the demodulator of the communication receiver 150 used to receive the measured signals. This may be understood to correspond to “Cut 2” in any of Figure 4 and Figure 6, which will be described later.

The obtaining of the plurality of estimations of second radio signals based on measurements may have been performed when interference may be low, in order to simplify the process. In some embodiments, the second radio signals may have a measured power above a second threshold.

Any of the first threshold and the second threshold may be a received power threshold. The power threshold may be a reference signal received power threshold.

In some embodiments, the first threshold and the second threshold may be the same threshold.

In some examples, the plurality of estimations of the second radio signals may be computed by leveraging the output of a subsequent functionality of the receiver 150, e.g., the soft bits, and reverting back the functions of the receiver 150, until the input of the relevant AI/ML model to be trained, e.g., by going backwards in the computational graph of the receiver 150. For example, the received useful data signals in Cut 2.2 of Figure 6 may be computed by leveraging, e.g., as the product between, the estimated channel and the transmitted data symbols computed based on the correctly decoded bits, e.g., after a successful Cyclic Redundancy Check (CRC).

By in this Action 301 obtaining the plurality of estimations of the second radio signals, the first node 111 may be enabled to obtain, e.g., estimate and/or store a set of AI/ML model inputs without or with limited noise and/or interference, relevant to links between the first cell 121 , e.g., sector and the devices served by the first cell 121 , e.g., the first device 131. The channel realizations may be well estimated because they may be captured in high SINR, where channel estimation may be easier.

Action 302

As will be explained in the next Action 303, according to embodiments herein, the first node 111 may use features in the method performed. A feature comprises a subset of the plurality of estimations of the second radio signals combined with one or more estimations of noise and/or the first radio signals.

As will be explained in the next Action 303, the first node 111 may also use, for every feature, a respective target of the feature. A target may be understood as the ground truth, that is, the ideal expected output of the model.

In this Action 302, the first node 111 may obtain the respective targets of the features.

The respective target of the feature may be one of: a) obtained by applying one or more functions to one of the estimations in the subset, and b) one of the estimations in the subset.

The functions may be for example, channel estimation, equalization, soft demapping, decoding, coding.

As explained before, the obtaining in this Action 302 may be understood as receiving. The receiving may be, in some examples, from another node where the calculation of the features or the targets may have been performed. In other examples, receiving may be understood as fetching, e.g., from another node or memory in the computer system 100, where the calculations may be stored. Yet in other examples, the first node 111 may obtain the respective targets of the features by estimating, performing the calculations, derivations or similar, itself In some examples, obtaining may comprise storing.

In some embodiments, the respective targets of the features may be obtained based on a type of estimation the mathematical model may be to be trained for. That is, the method to estimate the labels may depend on the AI/ML model functionality for which the data collection may be being performed. In these embodiments, as illustrative examples, one of the following options may apply. According to a first option, with the proviso the mathematical model may be for joint channel estimation, equalization and soft demapping, the respective targets of the features may correspond to coded bits. In such embodiments, the coded bits may be obtained by re-encoding the successfully decoded information bits, e.g., after a successful CRC check.

According to a second option, with the proviso the mathematical model may be for channel estimation, the respective targets of the features may correspond to channel realizations estimated using a demapper based on a set of measured reference signals. In this case, the respective targets or labels may correspond to the actual channel realization, which may be estimated using a channel estimator based on a set of measured reference signals. Example of such reference signals in NR may be DMRSs, SRSs, or CSI-RSs. In some embodiments, PUSCH/Physical Downlink Shared Channel (PDSCH) data may be used as a reference signal, e.g., through: 1) calculating the coded bits after the information bits may have been successfully decoded, e.g., re-encoding, 2) mapping the coded bits to the corresponding modulation symbols, and 3) calculating the ratio between the received signal after interference and noise removal and such modulation symbols.

According to the foregoing, in this Action 302, the first node 111 may obtain a set of labels relevant to links between sector and terminals that is, between the one or more devices 130 operating, or having operated, in the communications system 101 , and the one or more cells 120 in the communications system 101 , e.g., a set of channel realizations relevant to links between sector and terminals in the case of an AI/ML model for channel estimation, or the correctly decoded data bits in the case of an AI/ML model for joint channel estimation, equalization, and soft demapping.

It may be noted that the respective targets of the features, that is, the the labels, may be computed in slots where interference may be relatively low, may be understood to guarantee good quality of the available estimates, while simplifying the process.

In some examples, the plurality of estimations of the second radio signals obtained in Action 301 may be identical to the labels obtained in this Action 302. For example, this may be the case of an AI/ML model for channel estimation, whose inputs may be the channels realizations including noise and interference. Action 303

In this Action 303, the first node 111 combines the plurality of estimations of noise and/or the first radio signals with one or more estimations of the plurality of estimations of the second radio signals to generate a plurality of pairs. Each pair in the plurality of pairs comprises a feature. The feature comprises a subset of the plurality of estimations of the second radio signals combined with one or more estimations of noise and/or the first radio signals. In other words, the features may be created by combining the noise and interference realizations obtained in Action 301 with the noiseless/interference-less AI/ML model inputs obtained in Action 301.

The pair also comprises a respective target of the feature. The respective target of the feature is one of: a) obtained by applying the one or more functions to one of the estimations in said subset, and b) one of the estimations in the subset.

In some embodiments, the subset of the plurality of estimations of the second radio signals may correspond to transmitted scheduling information relating to the first cell 121 intended for multiple devices.

In other embodiments, the subset of the plurality of estimations of the second radio signals may correspond to transmitted scheduling information relating to the first cell 121 intended for a single device.

The combining in this Action 303 may be performed based on at least one of: a) one of: i) a linear function, and ii) a non-linear combination, and b) matching a target signal to interference and noise ratio.

In accordance with option a.i), in some embodiments, combining may be understood as performing a linear combination. It may be noted that a linear combination may comprise multiplying each of the components with a scalar and adding.

In other embodiments, in accordance with option a.ii), combining may be understood as performing a non-linear combination. It may be noted that a non-linear combination may comprise applying a non-linear transformation to each of the components and adding, e.g., to characterize the impact of non-linear effects, such as hardware impairments and/or nonlinearities. In yet other embodiments, combining may be understood as comprising both linear and non-linear combinations.

In some embodiments, in accordance with option b), the combination may be performed so that SINR may be matched to a target SINR. This may be understood to allow ensuring a good SINR distribution in the training data.

In some embodiments the Automatic Gain Control (AGC) state of the receiver 150 at the point of estimating the channel and interference and/or noise parts may be taken into account in the combining, which may be understood to mean, it may need to be stored. AGC may be understood to amplify or attenuate the received signal to maintain the output signal power within a suitable range. By taking into account the AGC state of the receiver 150 at the point of estimating the channel and interference and/or noise parts in the combining, the impact of the AGC may be removed, which may facilitate training.

Generalize the method to other receivers with a different AGC.

By combining the plurality of estimations of noise and/or the first radio signals with one or more estimations of the plurality of estimations of the second radio signals to generate the plurality of pairs of features and their respective targets in this Action 303, the first node 111 may enable to then train, using machine learning, a mathematical model with the combined pairs, so that the mathematical model may learn to identify the target from an inputted feature. The first node 111 may enable that training data for a ML-based mathematical model may be constructed so that important and challenging low SINR cases, the importance of which will be illustrated in Figure 5, may be guaranteed to be sufficiently represented. The channel realizations may be well estimated because they may be captured in high SINR, where channel estimation may be easier. The high interference/noise cases may also be captured. The training data set may be enabled to be larger because any combination of noise/interference and channel data may be used to form a training sample, so that data augmentation may be facilitated. All these advantages may be understood to lead to a better ML solution.

Action 304

In this Action 304, the first node 111 initiates training, using machine learning, of a mathematical model of estimations of radio signals in radio time-frequency resources. The training is performed using, for the obtained plurality of pairs, the feature as input and the respective target as ground truth.

Initiating training may be understood as starting the training itself, or triggering, enabling, facilitating another node, e.g., the second node 112, to perform the training.

Training may be understood to be the initial step in machine learning through which a working model may be approximated that may then be validated and implemented in a deployment. During such training, the mathematical model, e.g., a supervised ML model, may learn to map between features and targets from a dataset and may approximate an underlying function. In this process, the mathematical model may learn the best weights and biases using a gradient descent which may minimize an approximation loss through empirical risk minimization.

The training of the mathematical model that may be performed in some examples of this Action 304 may comprise use of training data samples comprising the features and the labels obtained, e.g., computed, in Action 302 and combined in Action 303, to train and/or update an AI/ML model. In some embodiments, e.g., wherein the first node 111 initiating the training may comprise the first node 111 performing the training itself, the training of the mathematical model may be performed until reaching a desired level of accuracy.

The training of the mathematical model may be performed according to best practice machine learning model training.

By initiating the training of the mathematical model in this Action 304, the first node 111 then enable that, once the mathematical model may have been trained, the mathematical model may be used to, e.g., perform the function for which the mathematical model may have been trained, e.g., channel estimation, or channel estimation, equalization, and soft demapping, on any input received signal, with a high level of accuracy. As will be shown with an example in Figure 5, ML based channel estimation has been shown to outperform the legacy channel estimator.

Action 305

In this Action 305, the first node 111 may send a first indication of the trained mathematical model to the second node 112 operating in the computer system 100.

By sending the first indication of the trained mathematical model to the second node 112 in this Action 305, the first node 111 may enable that the second node 112 may use the trained mathematical model to perform the function for which it may have been trained, e.g., channel estimation, or channel estimation, equalization, and soft demapping, on any input received signal, with a high level of accuracy.

Action 306

In this Action 306, the first node 111 may use the trained mathematical model to estimate another radio signal in a third set of time frequency resources.

By using the trained mathematical model to estimated the another radio signal in this Action 306, the first node 111 may be enabled to itself use the trained mathematical model to perform the function for which it may have been trained, e.g., channel estimation, or channel estimation, equalization, and soft demapping, on any input received signal, with a high level of accuracy.

Figure 4 is a schematic diagram depicting an illustration of the estimation/storage of a set of noise/interference realizations relevant to the sector when no user is scheduled, as performed according to Action 301 using the communication receiver 150 to receive the measured signals. The cross to the left of the diagram indicates that, in some embodiments the obtaining of the plurality of estimations of noise and/or first radio signals may be performed when no user is scheduled. For instance, the noise/interference realizations may be measured in Cut 2.1 in Figure 2-2 for the case of AI/ML-based channel estimation or in Cut 2 in Figure 2- 2 for joint channel estimation, equalization, and soft demapping. At cut 2.1 , if present the actual channel realization, may be estimated using a channel estimator based on a set of measured reference signals, such as e.g., DMRSs, SRSs, or CSI-RSs in resource elements. At cut 2.2, the actual transmitted symbols, may be estimated using at least resource elements with data, such as PUSCH/ PDSCH, used as a reference signal, e.g., through: 1) calculating the coded bits after the information bits may have been successfully decoded, e.g., reencoding, 2) mapping the coded bits to the corresponding modulation symbols, and 3) calculating the ratio between the received signal after interference and noise removal and such modulation symbols. The components of the receiver depicted may be understood to be otherwise similar to those already described in relation to Figure 1.

Figure 5 is a schematic diagram depicting a comparison plot showing performance gain using ML for channel estimation in experimental results. The solid black circles show data obtained with an ML based channel estimator, according to embodiments herein. The white circles show data obtained with the legacy (AIC) channel. The horizontal axis shows Signal to Noise Ratio (SNR) in dB. The vertical axis shows a logarithmic function of PUSCH throughput (tp) in Megabits per second (Mbps). As indicated on the top of the graph, the channel model corresponds to a clustered delay line A (CDL-A) with a delay spread of 100ns delay spread (DS), 10 deg Zenith of departure (ZoD) spread and 80 deg ZoD. As depicted in the Figure, the ML based channel estimator according to embodiments herein outperforms the legacy AIC based channel estimator. It may be noted that performance of the legacy (AIC) channel is significantly lower than the ML model for lower SNR, but performance is rather similar for high SNR. This shows that channel estimation may be easy for high SINR and hard for low SINR.

Figure 6 is a schematic diagram depicting an illustrative example for the estimation of part of the AI/ML inputs without or with limited noise/interference, as performed according to Action 301 using the communication receiver 150 to receive the measured signals, for the case of an AI/ML model for joint channel estimation, equalization, and soft demapping. In some examples, the input of the relevant AI/ML model inputs without or with very limited noise/interference may be computed by leveraging the output of a subsequent functionality of the receiver 150, e.g., the soft bits, and reverting back the functions of the receiver 150, until the input of the relevant AI/ML model to be trained, e.g., by going backwards in the computational graph of the receiver 150. For example, the received useful data signals (ftusefui Xusefui) in Cut 2.2 may be computed by leveraging, e.g., as the product between, the estimated channel and the transmitted data symbols (x_USefui) computed based on the correctly decoded bits, e.g., after a successful CRC. In Figure 6, y₂₂ may be understood to represent the signal at the input of the equalizer, assuming that an intended signal is present (ftusefuiXusefui), that one interferer is present (/7_int x_int) and that there is noise (n). x_usefui may be understood to represent transmitted symbols intended to the receiver; any term with the “A” sign may be understood to represent estimate. /7_usefui may be understood to represent the wireless channel response between the device that transmitted a signal intended to be decoded by the receiver 150. h may be understood to represent the wireless channel. x_int may be understood to represent interfering signals. h may be understood to represent the interfering channels, e.g., between the interferers and the receiver 150. It may be noted that, in practice, there may be several of them, that is, an addition. In the figure, only one is considered for simplicity. b_SOft may be understood to represent the estimated soft bits, b may be understood to represent the decoded bits. y_Ste_P3 may be understood to represent one way to estimate the received useful signal. The dashed lines may be understood to represent how the receiver 150 may estimate the received useful signal (/7usefuix_usefui), that is, yste_P3- Particularly, it shows how the receiver 150 may reencode the correctly decoded bits '£>' and perform symbol mapping to determine the transmitted signal x_USefui, and how the estimated channel may be used to calculate yste_P3- The components of the receiver depicted may be understood to be otherwise similar to those already described in relation to Figure 1.

Embodiments of a computer-implemented method, performed by the second node 1 12, will now be described with reference to the flowchart depicted in Figure 7. The method is for handling estimations of radio signals. The second node 1 12 operates in the computer system 100.

Several embodiments are comprised herein. In some embodiments all the actions may be performed. In some embodiments, some embodiments of the actions may be optional. It should be noted that the examples herein are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. The detailed description of some of the following corresponds to the same references provided above, in relation to the actions described for the first node 1 1 1 , and will thus not be repeated here. For example, the in some embodiments, any of the estimations, that is, of any of the two pluralities of estimations, may have been performed using synthetic channel realizations.

Action 701

In this Action 701 , the second node 1 12 obtains, directly or indirectly, for a fourth set of time frequency resources configured for use for communications between the second device 132 operating, or having operated, in the communications system 101 , and the second cell 122 in the communications system 101 one or more measurements. The fourth set of time frequency resources may be understood as a new set of time-frequency resources.

Action 702

In this Action 702, the second node 112 receives the first indication from the first node 111 operating in the computer system 100. The first indication indicates the mathematical model of estimations of radio signals in radio time-frequency resources described in relation to Figure 3, in Action 304, trained, using machine learning. The training has been performed using, for the obtained plurality of pairs described in relation to Figure 3 in Action 303, the feature as input and the respective target as ground truth.

Action 703

In this Action 703, the second node 112 uses the trained mathematical model to estimate a further radio signal in the fourth set of time frequency resources. The further radio signal may be understood as a new radio signal. That is, the second node 112 may be understood as a node executing the trained mathematical model to make inferences with it.

Any of the features described in Figure 3 leading to the training of the mathematical model, or any other features described in relation to Figure 3, may be understood to equally apply to the method performed by the second node 112.

As a summarized overview of the foregoing, embodiments herein may be understood to creating training data by combining the separately estimated useful signals with channel and noise/interference realizations that may be obtained independently.

Certain embodiments herein may provide one or more of the following technical advantage(s). Embodiments herein may be understood to enable that training data may be constructed so that the important and challenging low SINR cases may be guaranteed to be sufficiently represented. The channel realizations may be well estimated because they may be captured in high SINR, where channel estimation may be easier. The high interference/noise cases may also be captured. The training data set may be larger because any combination of noise/interference and channel data may be used to form a training sample, e.g., data augmentation may be facilitated. All these advantages may be expected to lead to a better ML solution.

Figure 8 depicts an example of the arrangement that the first node 111 may comprise to perform the method described in Figure 3 and/or Figures 4-6. The first node 111 may be understood to be for handling the estimations of radio signals. The first node 111 is configured to operate in the computer system 100.

Several embodiments are comprised herein. It should be noted that the examples herein are not mutually exclusive. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. The detailed description of some of the following corresponds to the same references provided above, in relation to the actions described for the first node 111 , and will thus not be repeated here. For example, the in some embodiments, any of the estimations, that is, of any of the two pluralities of estimations, may be configured to have been performed using synthetic channel realizations.

The first node 111 is configured to obtain, directly or indirectly, for the plurality of radio time-frequency resources configured for use for communications between the one or more devices 130 being configured to be operating, or configured to have operated, in the communications system 101 , and the one or more cells 120 in the communications system 101 the following. The first node 111 is configured to obtain the plurality of estimations of noise and/or first radio signals in the first set of time-frequency resources configured for use in the first cell 121 of the one or more cells 120, wherein the first radio signals are configured to lack corresponding transmitted scheduling information relating to the first cell 121 . The first node 111 is also configured to obtain the plurality of estimations of the second radio signals in the second set of time-frequency resources configured for use in the first cell 121 . The second radio signals are configured to be generated due to transmitted scheduling information relating to the first cell 121.

The first node 111 is also configured to combine the plurality of estimations of noise and/or the first radio signals with the plurality estimations of the plurality of estimations of the second radio signals to generate the plurality of pairs. Each pair in the plurality of pairs is configured to comprise: i) the feature configured to comprise the subset of the plurality of estimations of the second radio signals combined with the one or more estimations of noise and/or the first radio signals, and ii) the respective target of the feature. The respective target of the feature is configured to be one of: a) obtained by applying one or more functions to one of the estimations in the subset, and ii) the one of the estimations in the subset.

The first node 111 is further configured to initiate training, using machine learning, of the mathematical model of estimations of radio signals in radio time-frequency resources. The training is configured to be performed using, for the plurality of pairs configured to be obtained, the feature as input and the respective target as ground truth. In some embodiments, the subset of the plurality of estimations of the second radio signals may be configured to correspond to transmitted scheduling information relating to the first cell 121 intended for multiple devices.

In some embodiments, the subset of the plurality of estimations of the second radio signals may be configured to correspond to transmitted scheduling information relating to the first cell 121 intended for a single device.

In some embodiments, the training of the mathematical model may be configured to be performed until reaching the desired level of accuracy.

In some embodiments, the first node 111 may be further configured to at least one of the following.

In some embodiments, the first node 111 may be further configured to send the first indication of the trained mathematical model to the second node 112 configured to operate in the computer system 100.

In some embodiments, the first node 111 may be further configured to use the trained mathematical model to estimate another radio signal in the third set of time frequency resources.

In some embodiments, the obtaining of the plurality of estimations of noise and/or first radio signals may be configured to comprise at least one of: a) performing measurements of the noise and/or first radio signals when the first set of time-frequency resources lack corresponding transmitted scheduling information relating to the first cell 121 , b) removing intended signal from respective measured signals in the first set of time-frequency resources 155, and c) performing the measurements between a demodulator and a demapper of the communication receiver 150 used to receive the measured signals.

In some embodiments, the obtaining of the plurality of estimations of the second radio signals may be configured to be based on the measurements performed in one of: a) between the output of the demodulator and the input to the decoder of the communication receiver 150 used to receive the measured signals, and b) at the input of the channel estimator of the communication receiver 150 used to receive the measured signals, and c) at the output of the demodulator of the communication receiver 150 used to receive the measured signals.

In some embodiments, the first node 111 may be further configured to obtain the respective targets of the features.

In some embodiments, the respective targets of the features may be configured to be obtained based on the type of estimation the mathematical model may be configured to be trained for. In some of such embodiments, one of the following may apply: a) with the proviso the mathematical model may be configured to be for joint channel estimation, equalization and soft demapping, the respective targets of the features may be configured to correspond to coded bits, and b) with the proviso the mathematical model may be configured to be for channel estimation, the respective targets of the features may be configured to correspond to channel realizations configured to be estimated using the demapper based on the set of measured reference signals.

In some embodiments, the combining may be configured to be performed based on at least one of: a) one of: a) the linear function, and b) the non-linear combination, and b) matching the target signal to interference and noise ratio.

In some embodiments, at least one of the following may apply: a) the first node 111 may be configured to be the one of the one or more devices 130 or the radio network node 141 , 142 configured to serve one of the one or more devices 130, b) the first node 111 may be configured to be the one or the first device 131 in the first cell 131 and the first radio network node 141 configured to serve the first cell 121 , c) the first radio signals may be configured to have the measured power below the first threshold, d) the second radio signals may be configured to have the measured power above the second threshold, e) any of the first threshold and the second threshold may be configured to be the received power threshold, f) the power threshold may be configured to be the reference signal received power threshold, g) the first threshold and the second threshold may be configured to be configured to be the same threshold, and h) any of the estimations may be configured to be performed using synthetic channel realizations

The embodiments herein in the first node 111 may be implemented through one or more processors, such as a processing circuitry 801 in the first node 111 depicted in Figure 8, together with computer program code for performing the functions and actions of the embodiments herein. A processor, as used herein, may be understood to be a hardware component. The program code mentioned above may also be provided as a computer program product, for instance in the form of a data carrier carrying computer program code for performing the embodiments herein when being loaded into the first node 111. One such carrier may be in the form of a CD ROM disc. It is however feasible with other data carriers such as a memory stick. The computer program code may furthermore be provided as pure program code on a server and downloaded to the first node 111.

The first node 111 may further comprise a memory 802 comprising one or more memory units. The memory 802 is arranged to be used to store obtained information, store data, configurations, schedulings, and applications etc. to perform the methods herein when being executed in the first node 111.

In some embodiments, the first node 111 may receive information from, e.g., any of the devices in the plurality of first devices 110, such as the first device 131 and/or the second device 132, the receiver 150, the second node 112, any of the one or more devices 130, the radio network node 141 , 142, and/or another structure in the communications network 101 and/ or the computer system 100, through a receiving port 803. In some embodiments, the receiving port 803 may be, for example, connected to one or more antennas in first node 111. In other embodiments, the first node 111 may receive information from another structure in the communications network 100 through the receiving port 803. Since the receiving port 803 may be in communication with the processing circuitry 801 , the receiving port 803 may then send the received information to the processing circuitry 801 . The receiving port 803 may also be configured to receive other information.

The processing circuitry 801 in the first node 111 may be further configured to transmit or send information to e.g., any of the devices in the plurality of first devices 110, such as the first device 131 and/or the second device 132, the receiver 150, the second node 112, any of the one or more devices 130, the radio network node 141 , 142, and/or another structure in the communications network 101 and/ or the computer system 100, through a sending port 804, which may be in communication with the processing circuitry 801 , and the memory 802.

Those skilled in the art will also appreciate that the units comprised within the first node 111 described above as being configured to perform different actions, may refer to a combination of analog and digital circuits, and/or one or more processors configured with software and/or firmware, e.g., stored in memory, that, when executed by the one or more processors such as the processing circuitry 801 , perform as described above. One or more of these processors, as well as the other digital hardware, may be included in a single Application-Specific Integrated Circuit (ASIC), or several processors and various digital hardware may be distributed among several separate components, whether individually packaged or assembled into a System-on-a-Chip (SoC).

Also, in some embodiments, the different units comprised within the first node 111 described above as being configured to perform different actions described above may be implemented as one or more applications running on one or more processors such as the processing circuitry 801.

Thus, the methods according to the embodiments described herein for the first node 111 may be respectively implemented by means of a computer program 805 product, comprising instructions, i.e., software code portions, which, when executed on at least one processing circuitry 801 , cause the at least one processing circuitry 801 to carry out the actions described herein, as performed by the first node 111. The computer program 805 product may be stored on a computer-readable storage medium 806. The computer-readable storage medium 806, having stored thereon the computer program 805, may comprise instructions which, when executed on at least one processing circuitry 801 , cause the at least one processing circuitry 801 to carry out the actions described herein, as performed by the first node 111. In some embodiments, the computer-readable storage medium 806 may be a non-transitory computer- readable storage medium, such as a CD ROM disc, or a memory stick. In other embodiments, the computer program 805 product may be stored on a carrier containing the computer program 805 just described, wherein the carrier is one of an electronic signal, optical signal, radio signal, or the computer-readable storage medium 806, as described above.

The first node 111 may comprise a communication interface configured to facilitate, or an interface unit to facilitate, communications between the first node 111 and other nodes or devices, e.g. any of the devices in the plurality of first devices 110, such as the first device 131 and/or the second device 132, the receiver 150, the second node 112, any of the one or more devices 130, the radio network node 141 , 142, and/or another structure in the communications network 101 and/ or the computer system 100. The interface may, for example, include a transceiver configured to transmit and receive radio signals over an air interface in accordance with a suitable standard.

In other embodiments, the first node 111 may comprise a radio circuitry 807, which may comprise e.g., the receiving port 803 and the sending port 804.

The radio circuitry 807 may be configured to set up and maintain at least a wireless connection with any of the devices in the plurality of first devices 110, such as the first device 131 and/or the second device 132, the receiver 150, the receiver 150, the second node 112, any of the one or more devices 130, the radio network node 141 , 142, and/or another structure in the communications network 101 and/ or the computer system 100. Circuitry may be understood herein as a hardware component.

Hence, embodiments herein also relate to the first node 111 operative to operate in the communications network 100. The first node 111 may comprise the processing circuitry 801 and the memory 802, said memory 802 containing instructions executable by said processing circuitry 801 , whereby the first node 111 is further operative to perform the actions described herein in relation to the first node 111 , e.g., in Figure 3 and/or Figures 4-6, e.g., with an architecture as depicted with the non-limiting example of Figure 5.

Figure 9 depicts an example of the arrangement that the second node 112 may comprise to perform the method described in Figure 7. The second node 112 may be understood to be for handling estimations of radio signals. The second node 112 is configured to operate in the computer system 100.

Several embodiments are comprised herein. It should be noted that the examples herein are not mutually exclusive. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. The detailed description of some of the following corresponds to the same references provided above, in relation to the actions described for the second node 112, and will thus not be repeated here. For example, the in some embodiments, any of the estimations, that is, of any of the two pluralities of estimations, may be configured to have been performed using synthetic channel realizations.

The second node 112 is configured to obtain, directly or indirectly, for the fourth set of time frequency resources configured for use for communications between the second device 132 configured to operate, or configured to have operated, in the communications system 101 , and the second cell 122 in the communications system 101 the one or more measurements.

The second node 112 is also configured to receive the first indication from the first node 111 configured to operate in the computer system 100. The first indication is configured to indicate the mathematical model of estimations of radio signals in radio time-frequency resources described in relation to Figure 3, in Action 304, configured to be trained, using machine learning. The training is configured to have been performed using, for the plurality of pairs configured to be obtained as described in relation to Figure 3, in Action 304, the feature as input and the respective target as ground truth.

The second node 112 is further configured to use the trained mathematical model to estimate the further radio signal in the fourth set of time frequency resources.

In some embodiments, the model is configured to have been trained by the first node 111 , as further described in relation to Figure 3.

The embodiments herein in the second node 112 may be implemented through one or more processors, such as a processing circuitry 901 in the second node 112 depicted in Figure 9, together with computer program code for performing the functions and actions of the embodiments herein. A processor, as used herein, may be understood to be a hardware component. The program code mentioned above may also be provided as a computer program product, for instance in the form of a data carrier carrying computer program code for performing the embodiments herein when being loaded into the second node 112. One such carrier may be in the form of a CD ROM disc. It is however feasible with other data carriers such as a memory stick. The computer program code may furthermore be provided as pure program code on a server and downloaded to the second node 112.

The second node 112 may further comprise a memory 902 comprising one or more memory units. The memory 902 is arranged to be used to store obtained information, store data, configurations, schedulings, and applications etc. to perform the methods herein when being executed in the second node 112.

In some embodiments, the second node 112 may receive information from, e.g., the first node 111 , the receiver 150, any of the devices in the plurality of first devices 110, such as the first device 131 and/or the second device 132, any of the one or more devices 130, the radio network node 141 , 142, and/or another structure in the communications network 101 and/ or the computer system 100, through a receiving port 903. In some embodiments, the receiving port 903 may be, for example, connected to one or more antennas in second node 112. In other embodiments, the second node 112 may receive information from another structure in the communications network 100 through the receiving port 903. Since the receiving port 903 may be in communication with the processing circuitry 901 , the receiving port 903 may then send the received information to the processing circuitry 901 . The receiving port 903 may also be configured to receive other information.

The processing circuitry 901 in the second node 112 may be further configured to transmit or send information to e.g., the first node 111 , the receiver 150, any of the devices in the plurality of first devices 110, such as the first device 131 and/or the second device 132, any of the one or more devices 130, the radio network node 141 , 142, and/or another structure in the communications network 101 and/ or the computer system 100, through a sending port 904, which may be in communication with the processing circuitry 901 , and the memory 902.

Those skilled in the art will also appreciate that the units comprised within the second node 112 described above as being configured to perform different actions, may refer to a combination of analog and digital circuits, and/or one or more processors configured with software and/or firmware, e.g., stored in memory, that, when executed by the one or more processors such as the processing circuitry 901 , perform as described above. One or more of these processors, as well as the other digital hardware, may be included in a single Application-Specific Integrated Circuit (ASIC), or several processors and various digital hardware may be distributed among several separate components, whether individually packaged or assembled into a System-on-a-Chip (SoC).

Also, in some embodiments, the different units comprised within the second node 112 described above as being configured to perform different actions described above may be implemented as one or more applications running on one or more processors such as the processing circuitry 901.

Thus, the methods according to the embodiments described herein for the second node 112 may be respectively implemented by means of a computer program 905 product, comprising instructions, i.e., software code portions, which, when executed on at least one processing circuitry 901 , cause the at least one processing circuitry 901 to carry out the actions described herein, as performed by the second node 112. The computer program 905 product may be stored on a computer-readable storage medium 906. The computer- readable storage medium 906, having stored thereon the computer program 905, may comprise instructions which, when executed on at least one processing circuitry 901 , cause the at least one processing circuitry 901 to carry out the actions described herein, as performed by the second node 112. In some embodiments, the computer-readable storage medium 906 may be a non-transitory computer-readable storage medium, such as a CD ROM disc, or a memory stick. In other embodiments, the computer program 905 product may be stored on a carrier containing the computer program 905 just described, wherein the carrier is one of an electronic signal, optical signal, radio signal, or the computer-readable storage medium 906, as described above.

The second node 112 may comprise a communication interface configured to facilitate, or an interface unit to facilitate, communications between the second node 112 and other nodes or devices, e.g. the first node 111 , the receiver 150, any of the devices in the plurality of first devices 110, such as the first device 131 and/or the second device 132, any of the one or more devices 130, the radio network node 141 , 142, and/or another structure in the communications network 101 and/ or the computer system 100. The interface may, for example, include a transceiver configured to transmit and receive radio signals over an air interface in accordance with a suitable standard.

In other embodiments, the second node 112 may comprise a radio circuitry 907, which may comprise e.g., the receiving port 903 and the sending port 904.

The radio circuitry 907 may be configured to set up and maintain at least a wireless connection with the first node 111 , the receiver 150, any of the devices in the plurality of first devices 110, such as the first device 131 and/or the second device 132, any of the one or more devices 130, the radio network node 141 , 142, and/or another structure in the communications network 101 and/ or the computer system 100. Circuitry may be understood herein as a hardware component.

Hence, embodiments herein also relate to the second node 112 operative to operate in the communications network 100. The second node 112 may comprise the processing circuitry 901 and the memory 902, said memory 902 containing instructions executable by said processing circuitry 901 , whereby the second node 112 is further operative to perform the actions described herein in relation to the second node 112, e.g., in Figure 7.

Claims

CLAIMS:

1. A computer-implemented method, performed by a first node (111), for handling estimations of radio signals, the first node (111) operating in a computer system (100), the method comprising:

- obtaining (301), directly or indirectly, for a plurality of radio time-frequency resources configured for use for communications between one or more devices (130) operating, or having operated, in a communications system (101), and one or more cells (120) in the communications system (101): a. a plurality of estimations of noise and/or first radio signals in a first set of time-frequency resources configured for use in a first cell (121) of the one or more cells (120), wherein the first radio signals lack corresponding transmitted scheduling information relating to the first cell (121), and b. a plurality of estimations of second radio signals in a second set of time-frequency resources configured for use in the first cell (121), wherein the second radio signals are generated due to transmitted scheduling information relating to the first cell (121), and

- combining (303) the plurality of estimations of noise and/or the first radio signals with the plurality estimations of the plurality of estimations of the second radio signals to generate a plurality of pairs, wherein each pair in the plurality of pairs comprises: a. a feature comprising a subset of the plurality of estimations of the second radio signals combined with one or more estimations of noise and/or the first radio signals, and b. a respective target of the feature, wherein the respective target of the feature is one of:

- obtained by applying one or more functions to one of the estimations in the subset, and

- one of the estimations in the subset, and

- initiating (304) training, using machine learning, of a mathematical model of estimations of radio signals in radio time-frequency resources, wherein the training is performed using, for the obtained plurality of pairs, the feature as input and the respective target as ground truth.

2. The method according to claim 1 , wherein the subset of the plurality of estimations of the second radio signals corresponds to transmitted scheduling information relating to the first cell (121) intended for multiple devices.

3. The method according to claim 1 , wherein the subset of the plurality of estimations of the second radio signals corresponds to transmitted scheduling information relating to the first cell (121) intended for a single device.

4. The method according to any of claims 1-3, wherein the training of the mathematical model is performed until reaching a desired level of accuracy.

5. The method according to claim 4, further comprising at least one of:

- sending (305) a first indication of the trained mathematical model to a second node (112) operating in the computer system (100), and

- using (306) the trained mathematical model to estimate another radio signal in a third set of time frequency resources.

6. The method according to any of claims 1-5, wherein the obtaining (301) of the plurality of estimations of noise and/or first radio signals comprises at least one of: a. performing measurements of the noise and/or first radio signals when the first set of time-frequency resources lack corresponding transmitted scheduling information relating to the first cell (121), b. removing intended signal from respective measured signals in the first set of time-frequency resources (155), and c. performing the measurements between a demodulator and a demapper of the communication receiver (150) used to receive the measured signals.

7. The method according to any of claims 1-6, wherein the obtaining (301) of the plurality of estimations of the second radio signals is based on measurements performed in one of: a. between an output of a demodulator and an input to a decoder of a communication receiver (150) used to receive the measured signals, and b. at an input of a channel estimator of the communication receiver (150) used to receive the measured signals, and c. at an output of the demodulator of the communication receiver (150) used to receive the measured signals.

8. The method according to any of claims 1-7, further comprising:

- obtaining (302) the respective targets of the features.

9. The method according to claim 8, wherein the respective targets of the features are obtained based on a type of estimation the mathematical model is to be trained for, and wherein one of: a. with the proviso the mathematical model is for joint channel estimation, equalization and soft demapping, the respective targets of the features correspond to coded bits, and b. with the proviso the mathematical model is for channel estimation, the respective targets of the features correspond to channel realizations estimated using a demapper based on a set of measured reference signals.

10. The method according to any of claims 1-9, wherein the combining (303) is performed based on at least one of: a. one of: a) a linear function, and b) a non-linear combination, and b. matching a target signal to interference and noise ratio.

11. The method according to any of claims 1-10, wherein at least one of: a. the first node (111) is one of the one or more devices (130) or a radio network node (141 , 142) serving one of the one or more devices (130), b. the first node (111) is one or a first device (131) in the first cell (131) and a first radio network node (141) serving the first cell (121), c. the first radio signals have a measured power below a first threshold, d. the second radio signals have a measured power above a second threshold, e. any of the first threshold and the second threshold is a received power threshold, f. the power threshold is a reference signal received power threshold, g. the first threshold and the second threshold are the same threshold, and h. any of the estimations are performed using synthetic channel realizations.

12. A computer-implemented method, performed by a second node (112), for handling estimations of radio signals, the second node (112) operating in a computer system (100), the method comprising: - obtaining (701), directly or indirectly, for a fourth set of time frequency resources configured for use for communications between a second device (132) operating, or having operated, in a communications system (101), and a second cell (122) in the communications system (101) one or more measurements,

- receiving (702) a first indication from a first node (111) operating in the computer system (100), the first indication indicating the mathematical model of estimations of radio signals in radio time-frequency resources of claim 1 , trained, using machine learning, wherein the training has been performed using, for the obtained plurality of pairs according to claim 1 , the feature as input and the respective target as ground truth, and

- using (703) the trained mathematical model to estimate a further radio signal in the fourth set of time frequency resources.

13. The method according to claim 12, wherein the model has been trained by a method according to any of claims 2-11.

14. A first node (111), for handling estimations of radio signals, the first node (111) being configured to operate in a computer system (100), the first node (111) being further configured to:

- obtain, directly or indirectly, for a plurality of radio time-frequency resources configured for use for communications between one or more devices (130) being configured to be operating, or configured to have operated, in a communications system (101), and one or more cells (120) in the communications system (101): a. a plurality of estimations of noise and/or first radio signals in a first set of time-frequency resources configured for use in a first cell (121) of the one or more cells (120), wherein the first radio signals are configured to lack corresponding transmitted scheduling information relating to the first cell (121), and b. a plurality of estimations of second radio signals in a second set of time-frequency resources configured for use in the first cell (121), wherein the second radio signals are configured to be generated due to transmitted scheduling information relating to the first cell (121), and

- combine the plurality of estimations of noise and/or the first radio signals with the plurality estimations of the plurality of estimations of the second radio signals to generate a plurality of pairs, wherein each pair in the plurality of pairs is configured to comprise: a. a feature configured to comprise a subset of the plurality of estimations of the second radio signals combined with one or more estimations of noise and/or the first radio signals, and b. a respective target of the feature, wherein the respective target of the feature is configured to be one of:

- one of the estimations in the subset, and

- initiate training, using machine learning, of a mathematical model of estimations of radio signals in radio time-frequency resources, wherein the training is configured to be performed using, for the plurality of pairs configured to be obtained, the feature as input and the respective target as ground truth.

15. The first node (111) according to claim 14, wherein the subset of the plurality of estimations of the second radio signals is configured to correspond to transmitted scheduling information relating to the first cell (121) intended for multiple devices.

16. The first node (111) according to claim 14, wherein the subset of the plurality of estimations of the second radio signals is configured to correspond to transmitted scheduling information relating to the first cell (121) intended for a single device.

17. The first node (111) according to any of claims 14-16, wherein the training of the mathematical model is configured to be performed until reaching a desired level of accuracy.

18. The first node (111) according to claim 17, being further configured to at least one of:

- send a first indication of the trained mathematical model to a second node (112) configured to operate in the computer system (100), and

- use the trained mathematical model to estimate another radio signal in a third set of time frequency resources.

19. The first node (111) according to any of claims 14-18, wherein the obtaining of the plurality of estimations of noise and/or first radio signals is configured to comprise at least one of: a. performing measurements of the noise and/or first radio signals when the first set of time-frequency resources lack corresponding transmitted scheduling information relating to the first cell (121), b. removing intended signal from respective measured signals in the first set of time-frequency resources (155), and c. performing the measurements between a demodulator and a demapper of the communication receiver (150) used to receive the measured signals.

20. The first node (111) according to any of claims 14-19, wherein the obtaining of the plurality of estimations of the second radio signals is configured to be based on measurements performed in one of: a. between an output of a demodulator and an input to a decoder of a communication receiver (150) used to receive the measured signals, and b. at an input of a channel estimator of the communication receiver (150) used to receive the measured signals, and c. at an output of the demodulator of the communication receiver (150) used to receive the measured signals.

21 . The first node (111) according to any of claims 14-20, being further configured to:

- obtain the respective targets of the features.

22. The first node (111) according to claim 21 , wherein the respective targets of the features are configured to be obtained based on a type of estimation the mathematical model is configured to be trained for, and wherein one of: a. with the proviso the mathematical model is configured to be for joint channel estimation, equalization and soft demapping, the respective targets of the features are configured to correspond to coded bits, and b. with the proviso the mathematical model is configured to be for channel estimation, the respective targets of the features are configured to correspond to channel realizations configured to be estimated using a demapper based on a set of measured reference signals.

23. The first node (111) according to any of claims 14-22, wherein the combining is configured to be performed based on at least one of: a. one of: a) a linear function, and b) a non-linear combination, and b. matching a target signal to interference and noise ratio.

24. The first node (111) according to any of claims 14-23, wherein at least one of: a. the first node (111) is configured to be one of the one or more devices (130) or a radio network node (141 , 142) configured to serve one of the one or more devices (130), b. the first node (111) is configured to be one or a first device (131) in the first cell (131) and a first radio network node (141) configured to serve the first cell (121), c. the first radio signals are configured to have a measured power below a first threshold, d. the second radio signals are configured to have a measured power above a second threshold, e. any of the first threshold and the second threshold is configured to be a received power threshold, f. the power threshold is configured to be a reference signal received power threshold, g. the first threshold and the second threshold are configured to be configured to be the same threshold, and h. any of the estimations are configured to be performed using synthetic channel realizations.

25. A second node (112), for handling estimations of radio signals, the second node (112) being configured to operate in a computer system (100), the second node (112) being further configured to:

- obtain, directly or indirectly, for a fourth set of time frequency resources configured for use for communications between a second device (132) configured to operate, or configured to have operated, in a communications system (101), and a second cell (122) in the communications system (101) one or more measurements,

- receive a first indication from a first node (111) configured to operate in the computer system (100), the first indication being configured to indicate the mathematical model of estimations of radio signals in radio time-frequency resources of claim 1 , configured to be trained, using machine learning, wherein the training is configured to have been performed using, for the plurality of pairs configured to be obtained according to claim 1 , the feature as input and the respective target as ground truth, and

- use the trained mathematical model to estimate a further radio signal in the fourth set of time frequency resources.

26. The second node (112) according to claim 25, wherein the model is configured to have been trained by the first node (111), configured according to any of claims 15-24.

27. A computer program (805), comprising instructions which, when executed on at least one processing circuitry (801), cause the at least one processing circuitry (801) to carry out the method according to any of claims 1-11.

28. A computer-readable storage medium (806), having stored thereon a computer program (805), comprising instructions which, when executed on at least one processing circuitry (801), cause the at least one processing circuitry (801) to carry out the method according to any of claims 1-11.

29. A computer program (905), comprising instructions which, when executed on at least one processing circuitry (901), cause the at least one processing circuitry (901) to carry out the method according to any of claims 12-13.

30. A computer-readable storage medium (906), having stored thereon a computer program (905), comprising instructions which, when executed on at least one processing circuitry (901), cause the at least one processing circuitry (901) to carry out the method according to any of claims 12-13.