Causal coding of channel state information

Abstract

A computer-implemented method performed by a first computing device (200, 18300) for causally encoding channel state information (CSI) is provided. The method comprises obtaining (1402) an indication of an error report of the CSI; retrieving (1404) the parameter data; obtaining (1406) a measurement value of the CSI data; and activating (1408) a machine learning (ML) model for CSI reporting. The ML model includes a causal relationship between measurements of the parameter data and the CSI data. The method further comprises encoding (1410) the measurement value of the CSI data to obtain encoded CSI measurement data; applying (1412) an ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; and transmitting (1414) the adapted CSI measurement data to the second computing device. A computer-implemented method performed by a second computing device and related methods and apparatus are also provided.

Description

Technical Field

[0001] The present disclosure generally relates to a computer-implemented method performed by a first computing device for causally encoding channel state information (CSI), and related methods and apparatus. Background Art

[0002] For potential gains in telecommunication systems, CSI can be sent between a base station (BS) and a user equipment (UE). For example, potential gains can be achieved in massive multiple-input multiple-output (MIMO) systems, switching with other antenna types, frequency division duplex (FDD) systems, point-to-point systems for obtaining downlink CSI from uplinks, distributed MIMO (D-MIMO) for obtaining CSI from a subset of access points for use with other remaining access points, etc. However, the overhead of bidirectional signaling for channel state estimation may sometimes offset the benefits of sending CSI. Some approaches may include using an autoencoder to compress CSI feedback. However, such an approach may lack domain adaptation based on, for example, environmental changes. See, e.g., CK Wen, WT Shih, and S. Jin, “Deep Learning for Massive MIMO CSI Feedback,” IEEE Wireless Communication Letters, vol. 7, no. 5, Oct. 2018, https: / / github.com / sydney222 / Python_CsiNet (“Wen”); Y. Sun, W. Xu, L. Liang, N. Wang, GY Li, and X. You, “A Lightweight Deep Network for Efficient CSI Feedback in Massive MIMO Systems,” arXiv:2105.10283, 2021 (“Sun”). Due to, for example, parameters related to the equipment of the BS vendor or the UE vendor, there may be a distribution shift between the training and test data, so that a machine learning (ML) model involving CSI feedback may be executed in an environment different from the environment included in the training of the ML model.

[0003] Addressing data sharing and ML model sharing based on, for example, feasibility and / or privacy constraints may be insufficient. In addition, excessive reference signaling may increase power consumption. For example, from the mobile network side, a reference signal that activates a power amplifier (PA) may not allow the PA to enter a sleep state. Summary of the invention

[0004] There are some challenges. While some approaches may include variational graph autoencoders, such approaches may not address CSI, domain adaptation, and / or energy conservation. There may be a lack of methods for causally encoding CSI using autoencoders that can adapt to changes in data distribution, thereby saving energy.

[0005] Certain aspects of the present disclosure and embodiments thereof may provide solutions to these or other challenges.

[0006] In various embodiments of the present disclosure, a computer-implemented method for causally encoding CSI is provided, performed by a first computing device. The method includes: obtaining an indication of an erroneous report of CSI; retrieving parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters; obtaining a measurement value of the CSI data; and activating a machine learning (ML) model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measurement value of the CSI data. The method also includes encoding the measurement value of the CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; and sending the adapted CSI measurement data to a second computing device.

[0007] In other embodiments, a computer-implemented method for decoding causally encoded CSI performed by a second computing device is provided. The method includes receiving adapted CSI data from a first computing device, wherein the adapted CSI data is output from an ML model for CSI reporting, the ML model including causal relationships between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and measurement values ​​of the CSI data. The method also includes decoding the adapted CSI measurement data to obtain decoded CSI; and using the decoded CSI for resource allocation.

[0008] In other embodiments, a computer-implemented method for causally encoding and decoding CSI performed by a second computing device is provided. The method includes: detecting an erroneous report of CSI by a first computing device; retrieving parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters; and obtaining a measurement value of the CSI data from the first computing device. The method also includes activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measured CSI data. The method also includes encoding the measured CSI data to obtain encoded CSI measurement data; and applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

[0009] In other embodiments, a first computing device is provided. The first computing device is configured to causally encode CSI. The first computing device includes: a processing circuit; and at least one memory coupled to the processing circuit. The memory includes instructions that, when executed by the processing circuit, cause the first computing device to perform operations. These operations include: obtaining an indication of an erroneous report of CSI; retrieving parameter data, the parameter data including multiple specific parameters and multiple general parameters; obtaining a measurement value of CSI data; and activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measurement value of the CSI data. The operations also include encoding the measurement value of the CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; and sending the adapted CSI measurement data to a second computing device.

[0010] In other embodiments, a first computing device configured to causally encode CSI is provided. The first computing device is adapted to perform operations. The operations include: obtaining an indication of an erroneous report of CSI; retrieving parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters; obtaining a measurement value of CSI data; and activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measurement value of CSI data. The operations also include encoding the measurement value of CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; and sending the adapted CSI measurement data to a second computing device.

[0011] In other embodiments, a computer program including program code is provided to be executed by a processing circuit of a first computing device, the first computing device being configured to causally encode CSI. Execution of the program code causes the first computing device to perform operations. The operations include: obtaining an indication of an erroneous report of CSI; retrieving parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters; obtaining a measurement value of the CSI data; and activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measurement value of the CSI data. The operations also include encoding the measurement value of the CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; and sending the adapted CSI measurement data to a second computing device.

[0012] In other embodiments, a computer program product is provided that includes a non-transitory storage medium that includes program code to be executed by a processing circuit of a first computing device that is configured to causally encode CSI. Execution of the program code causes the first computing device to perform operations. The operations include: obtaining an indication of an erroneous report of CSI; retrieving parameter data that includes a plurality of specific parameters and a plurality of general parameters; obtaining a measurement of the CSI data; and activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measurement of the CSI data. The operations also include encoding the measurement of the CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; and sending the adapted CSI measurement data to a second computing device.

[0013] In other embodiments, a second computing device is provided. The second computing device is configured to decode causally encoded CSI. The second computing device includes: a processing circuit; and at least one memory coupled to the processing circuit. The memory includes instructions that, when executed by the processing circuit, cause the second computing device to perform operations. These operations include receiving adapted CSI data from a first computing device, wherein the adapted CSI data is output from an ML model for CSI reporting, the ML model including causal relationships between parameter data, the parameter data including multiple specific parameters and multiple general parameters and measurement values ​​of the CSI data. These operations also include decoding the adapted CSI measurement data to obtain decoded CSI; and using the decoded CSI for resource allocation.

[0014] In other embodiments, a second computing device configured to decode causally encoded CSI is provided. The second computing device is adapted to perform operations. The operations include receiving adapted CSI data from a first computing device, wherein the adapted CSI data is output from an ML model for CSI reporting, the ML model including causal relationships between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and measurement values ​​of the CSI data. The operations also include decoding the adapted CSI measurement data to obtain decoded CSI; and using the decoded CSI for resource allocation.

[0015] In other embodiments, a computer program including program code is provided to be executed by a processing circuit of a second computing device, the second computing device being configured to decode causally encoded CSI. Execution of the program code causes the second computing device to perform operations. The operations include receiving adapted CSI data from a first computing device, wherein the adapted CSI data is output from an ML model for CSI reporting, the ML model including causal relationships between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and measurement values ​​of the CSI data. The operations also include decoding the adapted CSI measurement data to obtain decoded CSI; and using the decoded CSI for resource allocation.

[0016] In other embodiments, a computer program product is provided that includes a non-transitory storage medium, the non-transitory storage medium including program code to be executed by a processing circuit of a second computing device, the second computing device being configured to decode causally encoded CSI. Execution of the program code causes the second computing device to perform operations. The operations include receiving adapted CSI data from a first computing device, wherein the adapted CSI data is output from an ML model for CSI reporting, the ML model including causal relationships between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and measurement values ​​of the CSI data. The operations also include decoding the adapted CSI measurement data to obtain decoded CSI; and using the decoded CSI for resource allocation.

[0017] In other embodiments, a second computing device is provided. The second computing device is configured to causally encode and decode CSI. The second computing device includes: a processing circuit; and at least one memory coupled to the processing circuit. The memory includes instructions that, when executed by the processing circuit, cause the second computing device to perform operations. These operations include: detecting an erroneous report of CSI of a first computing device; retrieving parameter data, the parameter data including multiple specific parameters and multiple general parameters; and obtaining a measurement value of CSI data from the first computing device. These operations also include activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measured CSI data. These operations also include encoding the measured CSI data to obtain encoded CSI measurement data; and applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

[0018] In other embodiments, a second computing device configured for causal encoding and decoding of CSI is provided. The second computing device is adapted to perform operations. The operations include: detecting an erroneous report of CSI of a first computing device; retrieving parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters; and obtaining a measurement value of CSI data from the first computing device. The operations also include activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measured CSI data. The operations also include encoding the measured CSI data to obtain encoded CSI measurement data; and applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

[0019] In other embodiments, a computer program including program code is provided to be executed by a processing circuit of a second computing device, the second computing device being configured to causally encode and decode CSI. Execution of the program code causes the second computing device to perform operations. The operations include: detecting an erroneous report of CSI of a first computing device; retrieving parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters; and obtaining a measurement value of CSI data from the first computing device. The operations also include activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measured CSI data. The operations also include encoding the measured CSI data to obtain encoded CSI measurement data; and applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

[0020] In other embodiments, a computer program product is provided that includes a non-transitory storage medium that includes program code to be executed by a processing circuit of a second computing device that is configured to causally encode and decode CSI. Execution of the program code causes the second computing device to perform operations. These operations include: detecting an erroneous report of CSI by a first computing device; retrieving parameter data that includes a plurality of specific parameters and a plurality of general parameters; and obtaining a measurement value of CSI data from the first computing device. These operations also include activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measured CSI data. These operations also include encoding the measured CSI data to obtain encoded CSI measurement data; and applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

[0021] Certain embodiments may provide one or more of the following technical advantages: Based on including causal-aware CSI coding, domain adaptation and / or energy saving may be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings illustrate certain non-limiting embodiments of the inventive concept, which are included to provide a further understanding of the present disclosure and are incorporated into and constitute a part of this application. In the drawings:

[0023] Figure 1 is a schematic diagram showing an overview of operations according to some embodiments of the present disclosure;

[0024] Figure 2 is a signaling diagram illustrating operations performed by a first computing device (eg, UE) for causally encoding CSI according to the present disclosure;

[0025] Figure 3 is a signaling diagram illustrating operations performed by a second computing device (e.g., a gNB) for causally encoding and / or decoding CSI according to some embodiments of the present disclosure;

[0026] Figure 4 is a block diagram illustrating a causal-aware CSI coding system according to some embodiments of the present disclosure;

[0027] Figure 5 is a block diagram illustrating an encoder for capturing causal relationships according to some embodiments of the present disclosure;

[0028] Figure 6 are images of sample CSI data from indoor and outdoor settings;

[0029] Figure 7 and Figure 8 are respectively training and test loss graphs for simulating an example embodiment of the present disclosure;

[0030] Fig. 9 A to Fig. 9 D is an image of a sample reconstruction from a variational autoencoder (VAE) simulated by an example embodiment;

[0031] Fig.10 A to Fig.10 F is a result image of VAE reconstruction in which the causal parameters in the simulation of the example embodiment are changed;

[0032] Fig.11 are a CSI image reconstructed in a simulation and an original CSI image of an example embodiment;

[0033] Fig.12 are a CSI image reconstructed by proportional intervention in a simulation and an original CSI image of an example embodiment;

[0034] Fig.13 are a CSI counterfactual image and an original CSI image reconstructed in a simulation of an example embodiment;

[0035] Fig.14 is a flowchart of the operation of a first computing device according to some embodiments of the present disclosure;

[0036] Fig.15 and Fig.16 is a flowchart of the operation of a second computing device according to some embodiments of the present disclosure;

[0037] Fig.17 is a block diagram of a network according to some embodiments;

[0038] Fig.18 is a block diagram of a computing device according to some embodiments of the present disclosure;

[0039] Fig.19 is a block diagram of a computing device according to some embodiments of the present disclosure; and

[0040] Fig. 20 is a block diagram of a virtualization environment according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0041] Hereinafter, the inventive concept will be described more fully with reference to the accompanying drawings, in which examples of embodiments of the inventive concept are shown. However, the inventive concept can be embodied in a variety of different forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided so that the disclosure will be comprehensive and complete, and the scope of the inventive concept is fully conveyed to those skilled in the art. It should also be noted that these embodiments are not mutually exclusive. A component from one embodiment can be assumed by default to be present in / used in another embodiment.

[0042] The following description presents various embodiments of the disclosed subject matter. These embodiments are presented as teaching examples and are not to be construed as limiting the scope of the disclosed subject matter. For example, some details of the embodiments may be modified, omitted or expanded without departing from the scope of the subject matter.

[0043] As used herein, the term "first computing device" refers to a device capable of, configured, arranged and / or operable to causally encode CSI. As further discussed herein, examples of the first computing device include, but are not limited to, a computer, a decentralized edge device, a decentralized edge server, a distributed collection of access points (e.g., base stations) cooperating via a central processing unit (CPU), and a UE. The UE may include, for example, a smart phone, a mobile phone, a cellular phone, a voice over IP (VoIP) phone, a wireless local loop phone, a desktop computer, a personal digital assistant (PDA), a wireless camera, a game console or device, a music storage device, a playback device, a wearable terminal device, a wireless endpoint, a mobile station, a tablet computer, a laptop computer, a laptop embedded device (LEE), a laptop mounted device (LME), a smart device, a wireless client premises equipment (CPE), a vehicle-mounted or vehicle-embedded / integrated wireless device, etc. Other examples include any UE identified by the Third Generation Partnership Project (3GPP), including a narrowband Internet of Things (NB-IoT) UE, a machine type communication (MTC) UE, and / or an enhanced MTC (eMTC) UE. For example, in D-MIMO, the first computing device may include a distributed set of access points (APs) cooperating via a CPU, and channel estimation may be performed at the AP or centrally at the CPU, where channel estimates from all APs in the distributed set may be combined for precoding or receive combining. In addition, machine learning (ML) models may be distributed across different APs, and one central entity may combine them.

[0044] As used herein, the term "second computing device" refers to a device capable of, configured, arranged and / or operable to causally encode and / or decode CSI. As further discussed herein, examples of the second computing device include, but are not limited to, a computer, a decentralized edge device, a decentralized edge server, a cloud node, a cloud server, and a centralized or distributed AP (e.g., base station) in a radio access network (RAN) (e.g., GNodeB (gNB), evolved NodeB (eNB), a core network node, an access point (Ap) (e.g., a radio access point), etc.).

[0045] As used herein, the term "parameter data" refers to parameter data provided by a vendor (e.g., UE vendor, gNB vendor, etc.) and includes, but is not limited to, vendor-specific parameter data and general / common parameter data, as further discussed herein.

[0046] As discussed above, there are certain challenges. Some methods using autoencoders may not be able to adapt to changes in parameter data due to vendor differences and / or data drift within the same vendor for CSI compression. In addition, existing methods may lack the ability to apply causal coding for CSI compression to adapt to different fields and / or save energy.

[0047] Certain aspects of the present disclosure and embodiments thereof may provide solutions to these or other challenges. In some embodiments, operations are provided including causal-aware CSI encoding that can adapt to changes in parameter data distributions and perform domain adaptation by applying an ML model to received data (e.g., by applying a causal layer in a neural network).

[0048] In some embodiments, the operations further include creating an ML model of the underlying constraints that generate the observed data including the CSI data. The ML model may be a structural causal model (SCM), wherein the ML model captures causal relationships between observed or endogenous variables from the CSI data and unobserved or exogenous variables from the parameter data (e.g., UE speed, frequency, etc. discussed further herein).

[0049] In some embodiments, the causal graph is represented by a directed acyclic graph (DAG), where nodes (e.g., vertices) correspond to endogenous variables and directed edges represent causal parent-child relationships. Endogenous variables can be input vectors of the channel H matrix, and exogenous variables can include parameter data (e.g., power thresholds, channel quality indicators, etc. discussed further herein).

[0050] Certain embodiments may provide one or more of the following technical advantages. Including causal-aware encoding may avoid the need to retrain the ML model to adapt to data distribution or environmental changes. In addition, only one ML model may need to be stored on the network side (e.g., at the BS) and the device side (e.g., at the UE). Therefore, power consumption and bandwidth may be reduced based on the reduction of signaling overhead.

[0051] Various embodiments of the present disclosure relate to operations performed by a first computing device (e.g., a UE) and / or a second computing device (e.g., a BS such as a gNB) for causally aware encoding of CSI, which can result in improved domain adaptability and / or interpretability. For example, during training and / or deployment time, there may be different parameter data (such as thresholds) and / or different environmental operating conditions (e.g., UE mobility, such as moving speed and direction relative to the gNB, line-of-sight or non-line-of-sight propagation, channel richness, channel correlation, power usage of the UE and / or gNB, etc.) between different UE or gNB vendors. Another example of parameter data that can affect CSI reporting is power constraints, such as adjustments to power amplifiers, which will be further discussed herein.

[0052] According to some embodiments, the ML model performs operations to model the underlying constraints that generate the observed data including the CSI. The ML model may be a structural causal model (SCM). The ML model (e.g., SCM) captures causal relationships (e.g., CSI dependence on UE speed, CSI dependence on frequency, etc.) and may also address hidden confounding factors. Therefore, when the basic causal structure is captured, domain adaptation may be obtained by matching the causal layers (e.g., in a neural network-based SCM). Therefore, these operations may be scalable.

[0053] Various embodiments of the present disclosure relate to the operation of a single-cell downlink massive MIMO system having N t >>1 transmit antenna and a single receiver antenna at the first computing device side (eg, at the UE side). An Orthogonal Frequency Division Multiplexing (OFDM) system may include N c subcarriers, so the received signal at the nth subcarrier can be given by the following formula: Among them, h n 、v n 、x n and z n Denote the channel vector, precoding vector, data bearing symbol and additive noise of the nth subcarrier respectively. In this formula, the second computing device (eg, BS) expects the first computing device (eg, UE) to return an estimate of the channel H matrix via a feedback link.

[0054] Since the overhead can be high, some approaches implement autoencoders. See, e.g., Wen; Sun. The encoder can take as input the channel matrix H and reconstruct the channel matrix The autoencoder can be trained in an unsupervised learning setting, where the parameter set is updated by an adaptive moment estimation (ADAM) algorithm. The loss function can be the mean squared error (MSE), which is calculated as the average of all training samples. The ADAM optimizer and MSE are examples, and other optimization and error calculation methods can be used.

[0055] Various embodiments of the present disclosure relate to operations performed by a first computing device and / or a second computing device that enhance an autoencoder based on the use of an ML model. The ML model can use structural causal equations or functional causal models (FCMs) to model observed variables or endogenous variables (e.g., CSI H) as functions of exogenous (unobserved) variables from parameter data.

[0056] Figure 1 is a schematic diagram showing an overview of operations according to some embodiments of the present disclosure. As shown, parameter data 102 and CSI data H from a channel data source 104 may be included in a channel data service 100. The CSI data H 104 is provided to an encoder 106 and a loss function 116. During training, an ML model 110 (e.g., an SCM as shown) is constructed that models how the data distribution changes as the parameter data 102 provided to the ML model 110 changes. The ML model 110 captures the causal relationship between the parameter data and the measured CSI data. Therefore, the ML model 110 can be vendor-independent based on the inclusion of parameter data, which can handle changes in the output of the encoder 106 that is automatically encoded that are specific to changes in the parameter data.

[0057] The ML model 110 may be trained using the same loss function 116 as the decoder 114. As input data, the encoder 106 uses, for example, CSI channel data H available from the channel data service 100. The CSI channel data H may include features indicative of the estimated channel quality. As further discussed herein, the CSI channel data H may be a three-dimensional tensor, where the dimensions correspond to the transmit antenna ports of the second computing device (e.g., a base station such as a gNB), the receive antenna ports of the first computing device (e.g., a UE), and the frequency (divided into subcarriers or subbands). The encoder 106 outputs a temporal latent space representation Y temp , which is converted by the ML model 110 into a latent space that is input to the decoder 114. temp In addition, the ML model 110 uses the parameter data 102 to transform Y temp Convert to Y.

[0058] During training, back propagation is performed through decoder back propagation 118 and encoder back propagation 120 using the loss function 116. The loss function 116 is provided to the decoder back propagation 118 to propagate the loss function to the various nodes of the autoencoder, calculate the contribution of each weight to the loss function, and adjust the weights accordingly by using gradient descent. These gradients are provided to the encoder back propagation 120 so that the gradients are back propagated through the autoencoder layers and the encoder weights and biases are updated 122.

[0059] As further discussed herein, the ML model 110 may be in a first computing device (eg, a UE) or a second computing device (eg, a gNB). The first computing device or the second computing device may also include an encoder 106, and the second computing device may include a decoder 114.

[0060] In some embodiments, to reduce computational complexity on the decoder side, the ML model 110 is used when the decoder 114 incorrectly decodes the CSI measurement data (e.g., always incorrectly decodes the CSI measurement data). The first computing device (e.g., UE) and / or the second computing device (e.g., gNB) may indirectly observe incorrect decoding of channel quality, such as abnormally high packet rates, low throughput, handing over the UE to a target cell despite reporting good channel quality, dropped calls, failure to establish a radio access bearer, etc. In some embodiments, these observations are triggers for the method.

[0061] Figure 2 is a signaling diagram illustrating operations performed by a first computing device 200 (eg, UE) for causally encoding CSI. Figure 2 In operation, the ML model 110 is part of the first computing device 200 . Figure 3 is a signaling diagram illustrating operations performed by a second computing device 202 (eg, a gNB) for causally encoding and / or decoding CSI. Figure 3 In operation, the ML model 110 is part of the second computing device 202. Figure 2 and Figure 3 In the example embodiment shown, the first computing device 200 and the second computing device 202 respectively maintain an encoder and a decoder, and the ML model 110 converts the latent space generated by the encoder into the expected input of the decoder. Note that when the first computing device 200 is a UE and the UE changes cells, the process is reset.

[0062] refer to Figure 2 , including an alternative embodiment in which the UE 200 obtains 204 erroneously reported CSI. In one embodiment, the UE 200 detects 206 an erroneous report of CSI. In another embodiment, the UE 200 receives 208 a report of erroneous CSI from the gNB 202.

[0063] In operation 210, the gNB 202 requests the UE 200 to retrieve parameter data. In operation 212, the UE 200 obtains a measurement value of the CSI data and enhances the measurement value into parameter data. In operation 214, the UE 200 activates the ML model 110 for CSI reporting.

[0064] In operation 216, gNB 202 requests UE 200 to obtain CSI measurement values ​​and reports the CSI measurement values ​​to gNB 202. In operation 218, UE 200 obtains the measurement values, and in operation 220, uses encoder 106 to compress the obtained measurement values ​​into a latent space representation Y tempIn operation 222, the UE 200 uses the ML model 110 to convert Y temp =Y, and in operation 224, a response including the CSI measurement value Y is sent to gNB 202. In operation 226, gNB 202 decodes the CSI information given Y. In operation 228, gNB 202 uses the CSI information to make resource allocations (e.g., Figure 2 Physical resource block (PRB) allocation as shown in the example embodiment of FIG.

[0065] refer to Figure 3 In operation 300, gNB 202 requests UE 200 to obtain CSI measurement values ​​and reports the CSI measurement values ​​to gNB 202. In operation 302, UE 200 obtains CSI measurement values ​​and compresses CSI measurement data into a latent space representation Y using encoder 106. temp In operation 304, UE 200 sends a response including CSI measurement value Y to gNB 202. In operation 306, gNB 202 monitors the behavior of UE 200 over time. In operation 308, gNB 202 detects erroneous reporting of CSI of UE 200. In operation 310, gNB 202 retrieves parameter data. In operation 312, UE 200 sends a request including enhancing the UE's channel data measurement values ​​to general and specific parameters.

[0066] In operation 314, the gNB 202 activates the ML model 110 for CSI reporting. In operation 316, the gNB 202 uses the ML model 110 to convert Y temp =Y, and in operation 318, decode the CSI information given Y. In operation 320, gNB 202 uses the CSI information to perform resource allocation (e.g., Figure 3 Physical resource block (PRB) allocation as shown in the example embodiment of FIG.

[0067] Figure 4 is a block diagram showing a causal-aware CSI coding system according to some embodiments of the present disclosure. Figure 4As shown in the example embodiment of , specific parameter data 400 is quantized and encoded as a one-hot encoded tensor. General parameter data 402 is quantized and encoded as a one-hot encoded tensor. Observation data 404 including CSI data can be collected from multiple suppliers. The specific parameter data 400 is input to the ML model, which is shown as a structural causal model based on a graph neural network 408. The specific parameter data 400 and the general parameter data 402 are input to the GNN 408 as corresponding tensors. The encoder 406 automatically encodes the CSI data 404, and the output of the encoder 406 is input to the GNN 408. The tensor 410 that changes dynamically at runtime is also input to the GNN 408. In response to these inputs, the GNN 408 outputs an adapted output.

[0068] The general parameter data of various embodiments of the present disclosure include but are not limited to: 1. The speed of the first computing device (e.g., UE), which may also be referred to as mobility. The speed may be determined by a network-based solution (e.g., a fifth generation (5G) positioning type approach using triangulation and / or beamforming); and / or the speed may be determined by the first computing device itself (e.g., using overhead techniques such as satellite positioning information (e.g., Global Positioning System (GPS))). Latitude and longitude indicators may be used. 2. Line-of-sight (LoS) or non-line-of-sight (NLOS) propagation, which can be a Boolean value, for example, indicating whether the wave propagates along a direct path, or whether there is diffraction, refraction, and / or reflection due to obstacles. 3. Spatial channel correlation (e.g., a measure of independence between adjacent antennas in a MIMO antenna array). Spatial channel correlation can be measured, for example, by the antenna correlation coefficient (ACC), which indicates mutual dependence. For example, the lower the ACC, the more independent the antennas, typically resulting in higher bandwidth. Far-field radiation patterns and S-parameter characteristics are examples of processes that can be used to measure ACC. 4. Subcarrier frequency, e.g., the frequency range of each subcarrier in a MIMO antenna. 5. Environmental conditions that can affect signal propagation, such as precipitation, temperature, etc. 6. Instantaneous Available Bandwidth (IBW), a measure of the maximum amount of spectrum that can be handled by the network, e.g. in KHz; and / or 7. The current state of the first computing device (e.g., "idle" or "active" state)

[0069] The general parameters can be one-hot encoded, where each general parameter is quantized to n levels. In the example embodiment, n is 3 and is input to the autoencoder module as an nx8 bit tensor. The general parameters and specific parameters discussed below can be input as tensors when building the causal relationship graph.

[0070] Specific parameter data of various embodiments of the present disclosure include but are not limited to: 1. Perceived interference, such as signal-to-noise ratio used as an indicator or a measure of signal quality, such as reference signal received power (RSRP), reference signal received quality (RSRQ), etc. 2. A first computing device (eg, UE) threshold for a specific vendor, such as an acceptable level of loss in channel quality information (CQI) reconstruction at the network vendor. 3. Vendor-specific second computing device (e.g., gNB) thresholds. 4. Available power at the first computing device (eg, UE), represented by at least one of a plurality of indicators, such as a battery charging state, a rate of discharge, etc. 5. The second computing device (e.g., gNB) nominal power parameters, such as power amplifier (e.g., single or multiple transmissions). 6. DC / DC input power of the radio equipment; and / or 7. First calculate the device (eg, UE) defined power class, for example, the industry defined power class of the device's maximum transmit power, and the maximum and minimum effective isotropic radiated power (EIRP) on New Radio (NR).

[0071] As discussed above, when constructing a causal graph, specific parameter data can be input as tensors.

[0072] Some embodiments include operations such as using a graph neural network (GNN) to build an ML model (e.g., SCM). The GNN can be an additional encoding step at the encoder that can adapt to changes in exogenous variables depending on parameter data (e.g., supplier or environmental changes). After building the GNN using the observed data, the GNN can answer two types of queries: (a) intervention queries, e.g., what would happen if one of the exogenous variables changed by a certain amount, and / or (b) Counterfactual queries, e.g., what will happen to a particular factual sample given a structural causal equation.

[0073] The mean squared error (MSE) metric can be used to build ML models (e.g., GNNs). Figure 5 is a block diagram of an encoder for capturing causal relationships according to some embodiments of the present disclosure. Figure 5As shown, CSI data 501 is input to an autoencoder 406 trained using ADAM optimization and MSE. The output of the autoencoder 406 is input to a trained GNN 408 that captures causal relationships. The trained GNN 408 outputs a potential sample 502.

[0074] ML model layers (e.g., SCM layers) can be trained in a federated learning setting, where data from multiple vendors or from different environmental conditions from the same vendor can incrementally improve the model. Online adaptation can also be included, which can be vendor-specific to changes in environmental conditions, can be federated across vendors, etc.

[0075] In some embodiments that include an SCM, counterfactuals are generated that can help make the SCM interpretable.

[0076] In a simulated example embodiment of the method, CSI data from Wen was used. The CSI data included data collected from two settings: (1) indoor data from a Pico cellular scenario in the 5.3 GHz band. The base station was located at the center of a 20 m square; and (2) outdoor data from a rural setting in the 300 MHz band. The base station was located at the center of a 400 m square. For both scenarios, the UEs were randomly located within the squares. There were 32 antennas at the base station and 1024 subcarriers were used.

[0077] The CSI data is used to generate a synthetic parameter dataset where the matrix can be transformed using scaling, changing orientation, changing position x and position y. In a deployment scenario, data can be more easily collected by changing settings such as power and recording the distance of the UE from the base station and the speed of the UE and the vendor of the UE and gNB as available metadata. Figure 6 Sample CSI data from an indoor setting and an outdoor setting are shown.

[0078] In an example embodiment, causal factors that generate a CSI dataset are selected, including indoor / outdoor setting labels, scale (similar to power in a real setting), orientation, and (x,y) position. The orientation and (x,y) position can be the position of an antenna array in a real setting. The example embodiment includes the following operations: (1) constructing a variational autoencoder (VAE) from underlying labels and images; (2) constructing an SCM using the VAE and underlying labels as input; and (3) demonstrating intervention and counterfactual analysis when changing labels.

[0079] In an actual deployment, if the supplier-to-supplier variation using an ML model (e.g., SCM) is captured, the method may include adapting the coding to the updated supplier. Although in the simulation, the SCM is formed manually, in an actual deployment, the ML model may also be learned using an algorithm for causal discovery.

[0080] The causal reasons included in this simulation are: Shape: 3 values ​​{Indoor 0, Indoor 1, Outdoor 0} - this selects three cases Scale: 6 values ​​linearly distributed between (0.5, 1) - scale and orientation are 0 in this simulation Orientation: 40 values ​​in (0,2π) - scale and orientation are 0 in this simulation Position X: 32 values ​​in (0,1) Position Y: 32 values ​​in (0,1)

[0081] Training of the VAE was performed on a graphics processing unit (GPU) with 200 epochs and the ADAM optimizer, and a learning rate of 1.0e-3. The test loss was calculated after every 5 training epochs. Figure 7 and Figure 8 The training and test loss plots for this simulation are shown separately.

[0082] Fig. 9 A to Fig. 9 D is the image reconstructed from the sample from VAE. The accuracy can be improved by increasing the number of training epochs.

[0083] Fig.10 A to Fig.10 F is the resulting image reconstructed by VAE with the causal parameters changed. Fig.10 A is the original image without changes: Tensor([[0.,2.,4.,38.,7.,5.]], device='cuda:0'), and Fig.10 B is Fig.10 A. Sample VAE reconstruction of image A. Fig.10 C is the original image with shape transformation: Tensor([[0.,1.,4.,38.,7.,5.]], device="cuda:0"), and Fig.10 D is Fig.10 Sample VAE reconstruction of image C. Fig.10 E is the original image with orientation changes: Tensor([[0.,1.,4.,1.,7.,5.]], device="cuda:0"), and Fig.10 F is Fig.10 Sample VAE reconstruction of image E.

[0084] In the example embodiment, next, the SCM is built, and Fig.11 , Fig.12 and Fig.13 Observation, intervention, and counterfactual examples are shown in . Fig.11 Includes original CSI images and CSI images reconstructed using SCM. Fig.12 Includes the original CSI image and the CSI image reconstructed by proportional intervention. Fig.13 Includes original CSI images and reconstructed CSI counterfactual images.

[0085] Thus, certain embodiments may provide one or more of the following technical advantages. Use of causal autoencoding may result in adaptation of different suppliers based on intervening queries. As shown in simulated example embodiments, with sufficient training, the method may be adapted to Gaussian noise, and may also be applicable to other distributions of random noise. Energy and power savings may be achieved since different ML models do not need to be stored for each supplier. Through adjustment, one ML model may be adapted based on operating conditions and supplier settings. Furthermore, the ML model may be interpretable via counterfactuals. Furthermore, by using federated learning, the method may adapt to new suppliers and / or different environmental settings.

[0086] Fig.14 is a first computing device 200, 18300 (using Fig.18 For example, the module can be stored in Fig.18 The modules may be stored in the memory 18304 of the first computing device, and these modules may provide instructions such that when the instructions of the modules are executed by the processing circuit 18302 of the corresponding first computing device, the processing circuit 18302 performs the corresponding operations of the flowchart.

[0087] refer to Fig.14 , a computer-implemented method for causally encoding CSI performed by a first computing device is provided. The method includes: obtaining (1402) an indication of an erroneous report of CSI; retrieving (1404) parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters; obtaining (1406) a measurement value of the CSI data; and activating an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measurement value of the CSI data. The method also includes encoding (1410) the measurement value of the CSI data to obtain encoded CSI measurement data; applying the ML model to (1412) the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; and sending (1414) the adapted CSI measurement data to a second computing device.

[0088] ML models may include SCMs.

[0089] In some embodiments, the ML model models the CSI measurement data as a function of multiple specific parameters and multiple general parameters.

[0090] The ML model may include a directed acyclic graph including a plurality of nodes corresponding to the CSI measurement data and a plurality of directed edges representing causal parent-child relationships.

[0091] The obtaining ( 1402 ) may include at least one of: (i) detecting an error report of the CSI, and (ii) receiving a message including the error report of the CSI from a second computing device.

[0092] The plurality of general parameters may include at least one or more of: (i) a speed of the first computing device, (ii) a location of the first computing device, (iii) a moving direction of the first computing device, (iv) line-of-sight or non-line-of-sight propagation of signals between the first computing device and the second computing device, (v) spatial channel correlation, (vi) one or more subcarrier frequencies, (vii) environmental conditions, (viii) available bandwidth, and (ix) a current state of the first computing device.

[0093] The plurality of specific parameters may include at least one or more of: (i) a measurement of perceived interference, (ii) a threshold for a first computing device of a particular vendor, (iii) a threshold for a second computing device of a particular vendor, (iv) available power at the first computing device, (v) a nominal power parameter for the second computing device, (vi) an input power to the second computing device, and (vii) a specified power level for the first computing device.

[0094] In some embodiments, the method further includes constructing (1400) an ML model using the GNN, a plurality of specific parameters, a plurality of general parameters, and the CSI data. The constructed ML model can model how the distribution of the CSI measurement data changes with changes in the specific parameters and the general parameters.

[0095] The construction (1400) may include encoding a plurality of specific parameters and a plurality of general parameters, respectively; inputting (i) the encoded plurality of specific parameters as tensors, (ii) the encoded plurality of general parameters as tensors, and (iii) CSI data to the GNN; and training the GNN using an adaptive algorithm and an error metric. The trained GNN may capture the causal relationship between the plurality of specific parameters, the plurality of general parameters, and the CSI data.

[0096] In some embodiments, the ML model is trained in a federated learning setting.

[0097] In some embodiments, the first computing device is a UE and the second computing device is a network node.

[0098] Fig.14 Various operations in the flowchart of may be optional for some embodiments of the first computing device and related methods. For example, the operation of block 1400 may be optional.

[0099] Fig.15 and Fig.16 is a second computing device 202, 19300 (using Fig.19 For example, the module can be stored in Fig.19 The modules may be stored in the memory 19304 of the second computing device, and these modules may provide instructions such that when the instructions of the modules are executed by the processing circuit 19302 of the corresponding second computing device, the processing circuit 19302 performs the corresponding operations of the flowchart.

[0100] refer to Fig.15 , a computer-implemented method for decoding causally encoded CSI performed by a second computing device is provided. The method includes receiving (1500) adapted CSI data from a first computing device, wherein the adapted CSI data is output from an ML model for CSI reporting, the ML model including causal relationships between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and measurement values ​​of the CSI data. The method also includes decoding (1502) the adapted CSI measurement data to obtain decoded CSI; and using (1504) the decoded CSI for resource allocation.

[0101] refer to Fig.16 , a computer-implemented method for causally encoding and decoding CSI, performed by a second computing device, is provided. The method includes: detecting (1602) an erroneous report of CSI by a first computing device; retrieving (1604) parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters; and obtaining (1606) a measurement value of the CSI data from the first computing device. The method also includes activating (1608) an ML model for CSI reporting. The ML model includes a causal relationship between the parameter data and the measured CSI data. The method also includes encoding (1610) the measured CSI data to obtain encoded CSI measurement data; and applying (1612) the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

[0102] In some embodiments, the method further comprises: decoding (1614) the adapted CSI measurement data to obtain decoded CSI; and using (1616) the decoded CSI for resource allocation.

[0103] ML models may include SCMs.

[0104] In some embodiments, the ML model models the CSI measurement data as a function of multiple specific parameters and multiple general parameters.

[0105] The ML model may include a directed acyclic graph including a plurality of nodes corresponding to the CSI measurement data and a plurality of directed edges representing causal parent-child relationships between the CSI measurement data and the parameter data.

[0106] The plurality of general parameters may include at least one or more of: (i) a speed of the first computing device, (ii) a location of the first computing device, (iii) a moving direction of the first computing device, (iv) line-of-sight or non-line-of-sight propagation of signals between the first computing device and the second computing device, (v) spatial channel correlation, (vi) one or more subcarrier frequencies, (vii) environmental conditions, (viii) available bandwidth, and (ix) a current state of the first computing device.

[0107] The plurality of specific parameters may include at least one or more of: (i) a measurement of perceived interference, (ii) a threshold for a first computing device of a particular vendor, (iii) a threshold for a second computing device of a particular vendor, (iv) available power at the first computing device, (v) a nominal power parameter for the second computing device, (vi) an input power to the second computing device, and (vii) a specified power level for the first computing device.

[0108] In some embodiments, the method further includes constructing (1600) an ML model using the GNN, a plurality of specific parameters, a plurality of general parameters, and the CSI data. The constructed ML model can model how the distribution of the CSI measurement data changes with changes in the specific parameters and the general parameters.

[0109] The construction (1600) may include encoding multiple specific parameters and multiple general parameters respectively; inputting (i) the encoded multiple specific parameters as tensors, (ii) the encoded multiple general parameters as tensors, and (iii) CSI data to the GNN; and training the GNN using an adaptive algorithm and an error indicator. The trained GNN can capture the causal relationship between the multiple specific parameters, the multiple general parameters, and the CSI data.

[0110] In some embodiments, the ML model is trained in a federated learning setting.

[0111] In some embodiments, the first computing device is a UE and the second computing device is a network node.

[0112] Fig.16Various operations in the flowchart of may be optional for some embodiments of the second computing device and related methods. For example, Fig.16 The operations of blocks 1600, 1614, and 1616 may be optional.

[0113] Example embodiments of the methods of the present disclosure may be implemented in networks including but not limited to telecommunications networks, such as Fig.17 As shown. Telecommunication network 17102 may include access network 17104 (e.g., RAN) and core network 17106 (which includes one or more core network nodes 17108). Access network 17104 may include one or more access nodes 17110A and 17110B, such as network nodes (e.g., base stations), or any other similar third generation partnership project (3GPP) access nodes or non-3GPP access points. Network node 17110 facilitates first computing devices 17112A to 17112D (e.g., UEs) (e.g., through and / or other first computing devices) to connect directly or indirectly to core network 17106 via one or more wireless connections.

[0114] Example wireless communications via wireless connections include sending and / or receiving wireless signals using electromagnetic waves, radio waves, infrared waves, and / or other types of signals suitable for transmitting information without using wiring, cables, or other material conductors. In addition, in different embodiments, the network may include any number of wired or wireless networks, network nodes, UEs, computing devices, and / or any other components or systems capable of facilitating or participating in data and / or signal communications (whether via wired or wireless connections). The network may include any type of communication, telecommunications, data, cellular, radio networks, and / or other similar types of systems, and / or be connected to them through interfaces.

[0115] As a whole, the network 17100 enables a connection between the first computing device 17112 and the second computing device 17110. In this sense, the network 17100 can be configured to operate according to predefined rules or procedures such as a specific standard, which includes but is not limited to: Global System for Mobile Communications (GSM) Universal Mobile Telecommunications System (UMTS); Long Term Evolution (LTE) and / or other suitable 2G, 3G, 4G, 5G standards, or any applicable future generation standards (e.g., 6G); Wireless Local Area Network (WLAN) standards, such as the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standard (WiFi); and / or any other suitable wireless communication standards, such as Worldwide Interoperability for Microwave Access (WiMax), Bluetooth, Z-Wave, Near Field Communication (NFC) ZigBee, LiFi, and / or any Low Power Wide Area Network (LPWAN) standards (e.g., LoRa and Sigfox).

[0116] In some examples, telecommunication network 17102 is a cellular network that implements 3GPP standardized features. Therefore, telecommunication network 17102 can support network slicing to provide different logical networks to different devices connected to the telecommunication network. For example, telecommunication network 17102 can provide ultra-reliable low-latency communication (URLLC) services to some first computing devices (e.g., UEs), while providing enhanced mobile broadband (eMBB) services to other first computing devices, and / or providing massive machine type communication (mMTC) / massive IoT services to yet another first computing device.

[0117] In some examples, network 17100 is not limited to including a RAN, but rather includes any network containing any programmable / configurable decentralized access points or network elements that also record data from performance measurement points in network 17100.

[0118] In some examples, the first computing device and / or the second computing device are configured as computers without an attached radio / baseband, etc.

[0119] The disclosed method is lightweight and therefore suitable for distributed node and cloud implementations. Various distributed processing options suitable for data sources, storage, computation, and coordination can be used. For example, data sampling can be performed at a node (e.g., a worker node), where data analysis, reasoning, ML model creation, ML model sharing, and CSI encoding are performed at a cloud server (e.g., a master server).

[0120] In some embodiments, the core network virtual network functions (VNFs) include, but are not limited to, network data analysis functions and operation support systems (OSS).

[0121] In some embodiments, implementations include implementations in an Open RAN (ORAN).

[0122] The method of the present disclosure may be performed by a first computing device (e.g., Fig.17 Any first computing device 17112A through 17112D (one or more of which may be generally referred to as first computing device 17112) or Fig.18 The computing device 18300) executes.

[0123] refer to Fig.18 As previously discussed, the first computing device may include Figure 2 The first computing device 20o or Fig.18The computing device 18300 of the present invention includes a device capable of, configured, arranged and / or operable to causally encode CSI. As previously discussed, examples of the first computing device include, but are not limited to, a computer, a decentralized edge device, a decentralized edge server, and a UE. In some embodiments, the first computing device 18300 includes a processing circuit 18302, which is operably coupled to a memory 18304, an ML model and an autoencoder (not shown), and / or any other components or any combination thereof. Some of the first computing devices may utilize Fig.18 All or a subset of the components shown. The degree of integration between the components of each first computing device may vary. In addition, some first computing devices may include multiple instances of components, such as multiple processors, memories, transceivers, transmitters, receivers, etc.

[0124] The processing circuit 18302 is configured to process instructions and data, and may be configured to implement any sequential state machine operable to execute instructions stored in the memory 18304 and / or the ML model as a machine-readable computer program. The processing circuit 18302 may be implemented as one or more hardware-implemented state machines (e.g., in discrete logic, field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), etc.); programmable logic and suitable firmware; one or more stored computer programs, general-purpose processors (e.g., microprocessors or digital signal processors (DSPs)) and appropriate software; or any combination of the above. For example, the processing circuit 18302 may include multiple central processing units (CPUs).

[0125] The memory 18304 and / or the ML model may be or be configured to include a memory, such as a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic disk, an optical disk, a hard disk, a removable tape, a flash drive, etc. In one example, the memory 18304 and / or the ML model include one or more applications, such as an operating system, a web browser application, a widget, a gadget engine, or other applications, and corresponding data. The memory 18304 and / or the ML model may store any of a variety of operating systems or a combination of operating systems used by the first computing device 18300.

[0126] The memory 18304 may be configured to include multiple physical drive units, such as a redundant array of independent disks (RAID), a flash memory, a USB flash drive, an external hard drive, a thumb drive, a pen drive, a key drive, a high-density digital versatile disk (HD-DVD) optical drive, a built-in hard drive, a Blu-ray optical drive, a holographic digital data storage (HDDS) optical drive, an external mini dual in-line memory module (DIMM), a synchronous dynamic random access memory (SDRAM), an external micro DIMM SDRAM, a smart card memory (e.g., a tamper-proof module in the form of a universal integrated circuit card (UICC), including one or more subscriber identity modules (SIMs), such as USIM and / or ISIM), other memories, or any combination thereof. The UICC may be, for example, an embedded UICC (eUICC), an integrated UICC (iUICC), or a removable UICC commonly referred to as a "SIM card". The memory 18304 and / or the ML model may allow the first computing device 18300 to access instructions, applications, etc. stored on a temporary or non-temporary memory medium to offload data or upload data. An article of manufacture (e.g., an article of manufacture that utilizes a network) may be tangibly embodied as or in memory 18304 and / or an ML model, which may be or include a device-readable storage medium.

[0127] The processing circuit 18302 can be configured to communicate with an access network or other network using a communication interface 18306. The communication interface may include one or more communication subsystems and may include an optional antenna or be communicatively coupled to an optional antenna. The communication interface 18306 may include one or more transceivers for communication (e.g., by communicating with one or more remote transceivers of another device capable of wireless communication (e.g., another first computing device or a second computing device, such as a network node)). Each transceiver may include a transmitter and / or receiver suitable for providing network communications (e.g., optical, electrical, frequency allocation, etc.). In addition, optional transmitters and receivers may be coupled to one or more optional antennas and may share circuit components, software, or firmware, or alternatively be implemented separately.

[0128] In the illustrated embodiment, the communication function of the communication interface 18306 may include cellular communication, Wi-Fi communication, LPWAN communication, data communication, voice communication, multimedia communication, short-range communication such as Bluetooth, near field communication, location-based communication (e.g., using a global positioning system (GPS) to determine location), another type of communication function, or any combination thereof. Communication may be implemented according to one or more communication protocols and / or standards (e.g., IEEE 802.11, code division multiple access (CDMA), wideband code division multiple access (WCDMA), GSM, LTE, new radio (NR), UMTS, WiMax, Ethernet, transmission control protocol / Internet protocol (TCP / IP), synchronous optical network (SONET), asynchronous transfer mode (ATM), QUIC, hypertext transfer protocol (HTTP), etc.).

[0129] Additional methods of the present disclosure may be performed by a second computing device (e.g., Fig.19 The second computing device 19300 or Figure 2 The second computing device 19300 may be executed by a second computing device 202). As previously discussed, the second computing device includes a device capable of, configured, arranged and / or operable to causally encode CSI and / or decode CSI. Examples of the second computing device include, but are not limited to, a computer, a decentralized edge device, a decentralized edge server, a cloud node, a cloud server, and a centralized or distributed BS in a RAN (e.g., a gNB, an eNB, a core network node, an AP (e.g., a radio access point), etc.). In some embodiments, the second computing device 19300 includes modules, ML models, autoencoders and / or decoders (not shown) that can be stored in the memory 19304, and these modules may provide instructions such that when the instructions of the modules are executed by Fig.19 When the processing circuit 19302 executes, the second computing device performs corresponding operations of the method according to various embodiments of the present disclosure.

[0130] Some second computing devices may utilize Fig.19 All or a subset of the components shown. The degree of integration between the components of each second computing device may vary. In addition, some second computing devices may include multiple instances of components, such as multiple processors, memories, transceivers, transmitters, receivers, etc.

[0131] The processing circuit 19302 is configured to process instructions and data, and may be configured to implement any sequential state machine operable to execute instructions stored in the memory 19304, ML model, autoencoder, and / or decoder as a machine-readable computer program. The processing circuit 19302 may be implemented as one or more hardware-implemented state machines (e.g., in discrete logic, field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), etc.); programmable logic and suitable firmware; one or more stored computer programs, general-purpose processors (e.g., microprocessors or digital signal processors (DSPs)) and appropriate software; or any combination of the above. For example, the processing circuit 19302 may include multiple central processing units (CPUs).

[0132] The memory 19304, ML model, autoencoder and / or decoder may be or be configured to include a memory, such as a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic disk, an optical disk, a hard disk, a removable tape, a flash drive, etc. In one example, the memory 19304, ML model, autoencoder and / or decoder include one or more applications, such as an operating system, a web browser application, a widget, a gadget engine, or other applications, and corresponding data. The memory 19304, ML model, autoencoder and / or decoder may store any of a variety of operating systems or a combination of operating systems used by the second computing device 19300.

[0133] The memory 19304, ML model, autoencoder and / or decoder may be configured to include multiple physical drive units, such as a redundant array of independent disks (RAID), a flash memory, a USB flash drive, an external hard drive, a thumb drive, a pen drive, a key drive, a high-density digital versatile disc (HD-DVD) optical drive, an internal hard drive, a Blu-ray disc drive, a holographic digital data storage (HDDS) optical drive, an external mini dual in-line memory module (DIMM), a synchronous dynamic random access memory (SDRAM), an external micro DIMM SDRAM, a smart card memory (e.g., a tamper-proof module in the form of a universal integrated circuit card (UICC), including one or more subscriber identity modules (SIMs), such as USIM and / or ISIM), other memories, or any combination thereof. The UICC may be, for example, an embedded UICC (eUICC), an integrated UICC (iUICC), or a removable UICC commonly referred to as a "SIM card". The memory 19304, ML model, autoencoder and / or decoder may allow the second computing device 19300 to access instructions, applications, etc. stored on a temporary or non-temporary memory medium to offload data or upload data. An article of manufacture (e.g., an article of manufacture utilizing a network) may be tangibly embodied as or in memory 19304, an ML model, an autoencoder, and / or a decoder, which may be or include a device-readable storage medium.

[0134] The processing circuit 19302 may be configured to communicate with an access network or other network using a communication interface 19306. The communication interface 19306 may include one or more communication subsystems and may include an optional antenna or be communicatively coupled to an optional antenna. The communication interface 19306 may include one or more transceivers for communication (e.g., by communicating with one or more remote transceivers of another device capable of wireless communication (e.g., a first computing device or another second computing device)). Each transceiver may include a transmitter and / or receiver suitable for providing network communications (e.g., optical, electrical, frequency allocation, etc.). In addition, optional transmitters and receivers may be coupled to one or more optional antennas and may share circuit components, software, or firmware, or alternatively be implemented separately.

[0135] In the illustrated embodiment, the communication functions of the communication interface 19306 may include cellular communication, Wi-Fi communication, LPWAN communication, data communication, voice communication, multimedia communication, short-range communication such as Bluetooth, near field communication, location-based communication (e.g., using GPS to determine location), another type of communication function, or any combination thereof. Communication may be implemented according to one or more communication protocols and / or standards (e.g., IEEE 802.11, CDMA, WCDMA, GSM, LTE, NR, UMTS, WiMax, Ethernet, TCP / IP, SONET, ATM, QUIC, HTTP, etc.).

[0136] Fig. 20 20500 in which the functions implemented by some embodiments may be virtualized. In this context, virtualization means creating a virtual version of an apparatus or device that may include a virtualized hardware platform, storage device, and network resources. As used herein, virtualization may be applied to any device or component thereof described herein, and relates to an implementation in which at least a portion of the functions are implemented as one or more virtual components. Some or all of the functions described herein may be implemented as virtual components executed by one or more virtual machines (VMs) implemented in one or more virtual environments 20500 hosted by one or more of a hardware node (e.g., a hardware computing device operated as a second computing device (e.g., a network node), a first computing device (e.g., a UE), a core network node, or a host). In addition, in an embodiment where a virtual node does not require a radio connection (e.g., a core network node or a host), the node may be fully virtualized.

[0137] Application 20502 (which may alternatively be referred to as a software instance, a virtual appliance, a network function, a virtual node, a virtual network function, etc.) runs in a virtualized environment Q400 to implement some features, functions and / or benefits of some embodiments disclosed herein.

[0138] The hardware 20504 includes processing circuitry, memory storing software and / or instructions that can be executed by the hardware processing circuitry, and / or other hardware devices described herein (such as network interfaces, input / output interfaces, etc.). The software can be executed by the processing circuitry to instantiate one or more virtualization layers 20506 (also referred to as hypervisors or virtual machine monitors (VMMs)), provide VMs 20508a and VMs 20508b (one or more of which can be generally referred to as VMs 20508), and / or perform any functions, features, and / or benefits described in connection with some embodiments described herein. The virtualization layer 20506 can present a virtual operating platform to the VMs 20508 that looks like network hardware.

[0139] VM 20508 includes virtual processing, virtual memory, virtual network or interface, and virtual storage, and can be run by corresponding virtualization layer 20506. Different embodiments of instances of virtual device 20502 can be implemented on one or more VMs 20508, and these embodiments can be implemented in different ways. In some contexts, virtualization of hardware is referred to as network function virtualization (NFV or VNF). NFV can be used to unify numerous network device types onto industry standard high-capacity server hardware, physical switches, and physical storage that can be located in data centers and customer premises equipment (CPE).

[0140] In the context of NFV, a VM 20508 can be a software implementation of a physical machine that runs programs as if they were executed on a physical, non-virtualized machine. Each VM 20508 and the portion of the hardware 20504 that executes the VM (whether it is hardware dedicated to the VM and / or hardware shared by the VM with other VMs) form a separate virtual network element. Still in the context of NFV, a virtual network function is responsible for handling a specific network function running in one or more VMs 20508 on the hardware 20504 and corresponds to an application 20502.

[0141] The hardware 20504 may be implemented in a standalone network node with general or specific components. The hardware 20504 may implement some functions via virtualization. Alternatively, the hardware 20504 may be part of a larger hardware cluster (e.g., in a data center or CPE), where many hardware nodes work together and are managed by management and orchestration 20510, which in particular oversees the lifecycle management of the application 20502. In some embodiments, the hardware 20504 is coupled to one or more radio units, each of which includes one or more transmitters and one or more receivers that may be coupled to one or more antennas. The radio unit may communicate directly with other hardware nodes via one or more appropriate network interfaces, and may be used in conjunction with virtual components to provide radio capabilities to virtual nodes (e.g., radio access nodes or base stations). In some embodiments, some signaling may be provided by using a control system 20512, which may alternatively be used for communication between hardware nodes and radio units.

[0142] Although the first computing device and the second computing device described herein may include a combination of the hardware components shown, other embodiments may include a first computing device and / or a second computing device having a different combination of components. It should be understood that these first computing devices and / or second computing devices may include any suitable combination of hardware and / or software required to perform the tasks, features, functions, and methods disclosed herein. The determination, calculation, acquisition, or similar operations described herein may be performed by a processing circuit, which may process information by, for example, converting the obtained information into other information, comparing the obtained information or the converted information with the information stored in the first computing device and / or the second computing device, and / or performing one or more operations based on the obtained information or the converted information, and making a determination based on the result of the processing. In addition, although the components are depicted as a single box located within a larger box or nested within multiple boxes, in fact, the first computing device and / or the second computing device may include multiple different physical components that constitute a single illustrated component, and the functions may be divided between separate components. For example, a communication interface may be configured to include any component described herein, and / or the functions of the components may be divided between a processing circuit and a communication interface. In another example, the non-computationally intensive functions of any such component may be implemented in software or firmware, and the computationally intensive functions may be implemented in hardware.

[0143] In some embodiments, some or all of the functions described herein may be provided by a processing circuit that executes instructions stored in a memory, which in some embodiments may be a computer program product in the form of a non-transitory computer-readable storage medium. In alternative embodiments, some or all of the functions may be provided by a processing circuit, for example, in a hardwired manner, without executing instructions stored on a separate or discrete device-readable storage medium. In any of these specific embodiments, the processing circuit may be configured to perform the described functions, regardless of whether instructions stored on a non-transitory computer-readable storage medium are executed. The benefits provided by such functionality are not limited to separate processing circuits or to other components of the first computing device and / or the second computing device, but are enjoyed as a whole by the first computing device and / or the second computing device and / or are generally enjoyed by end users and wireless networks.

[0144] In the above description of various embodiments of the present disclosure, it is to be understood that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the inventive concept. Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meanings as commonly understood by ordinary technicians in the field to which the inventive concept belongs. It will be further understood that terms (e.g., terms defined in general dictionaries) should be interpreted as having meanings consistent with their meanings in the context of this specification and the relevant technology, and not interpreted as ideal or overly superficial meanings, unless so explicitly defined herein.

[0145] When an element is referred to as being "connected," "coupled," "responsive," or variations thereof relative to another element, it may be directly connected, coupled to, or responsive to the other element, or there may be an intermediate element. In contrast, when an element is referred to as being "directly connected," "directly coupled," "directly responsive," or variations thereof relative to another element, there is no intermediate element. Throughout the text, similar reference numerals represent similar elements. In addition, "coupling," "connection," "responsiveness," or variations thereof used herein may include wireless coupling, connection, or response. As used herein, the singular forms "one," "an," and "the" are intended to also include plural forms, unless the context clearly indicates otherwise. For the sake of brevity and / or clarity, well-known functions or structures may not be described in detail. The term "and / or" includes any and all combinations of one or more items listed in association.

[0146] It will be understood that although the terms first, second, third, etc. can be used herein to describe various elements / operations, these elements / operations should not be limited by these terms. These terms are only used to distinguish an element / operation from another element / operation. Therefore, the first element / operation in some embodiments can be referred to as the second element / operation in other embodiments without departing from the teaching of the inventive concept. Throughout this specification, the same reference numerals or the same reference signs represent the same or similar elements.

[0147] As used herein, the terms "includes," "comprising," "containing," "covering," "consisting of," "included," "having," "having," "having," or variations thereof are open ended and include one or more recited features, elements, steps, components, or functions, but do not preclude the presence or addition of one or more other features, elements, steps, components, functions, or combinations thereof. In addition, as used herein, the commonly used abbreviation "eg," derived from the Latin phrase "exempligratia," may be used to introduce or specify one or more general examples of a previously mentioned item, and is not intended as a limitation of that item. The commonly used abbreviation "ie," derived from the Latin phrase "idest," may be used to specify a specific item of a more general reference.

[0148] This article describes example embodiments with reference to block diagrams and / or flowchart illustrations of computer-implemented methods, devices (systems and / or equipment), and / or computer program products. It should be understood that the blocks of the block diagrams and / or flowcharts and the combination of blocks of the block diagrams and / or flowcharts can be implemented by computer program instructions executed by one or more computer circuits. These computer program instructions can be provided to a processor circuit of a general-purpose computer circuit, a special-purpose computer circuit, and / or other programmable data processing circuit to produce a machine, so that the instructions executed by the processor of the computer and / or other programmable data processing device convert and control transistors, values ​​stored in memory locations, and other hardware components within such circuits to implement the functions / actions specified in the block diagrams and / or flowcharts, and thereby create a device (functional body) and / or structure for implementing the functions / actions specified in the block diagrams and / or flowcharts.

[0149] These computer program instructions may also be stored in a tangible computer-readable medium that can direct a computer or other programmable data processing device to act in a specific manner so that the instructions stored in the computer-readable medium produce an article of manufacture, the article of manufacture including instructions for implementing the functions / actions specified in one or more boxes of the block diagram and / or flow chart. Therefore, embodiments of the present invention may be implemented in hardware and / or in software (including firmware, stored software, microcode, etc.) running on a processor such as a digital signal processor, which may be collectively referred to as a "circuit," "module," or variations thereof.

[0150] It should also be noted that in some alternative embodiments, the function / action marked in the frame may not occur in the order marked in the flow chart. For example, depending on the function / action involved, the two frames shown in succession can actually be performed substantially simultaneously, or the frame can sometimes be performed in the opposite order. In addition, the function of a given frame of a flow chart and / or block diagram can be divided into multiple frames, and / or the function of two or more frames of a flow chart and / or block diagram can be integrated at least in part. Finally, without departing from the scope of the inventive concept, other frames can be added / inserted between the frames shown, and / or the frame / operation can be omitted. In addition, although some frames include arrows about the communication path for indicating the main direction of communication, it should be understood that communication can occur in the direction opposite to the arrow represented.

[0151] Under the premise of not departing from the principle of the inventive concept basically, many changes and modifications can be made to the embodiment.All these changes and modifications are intended to be included in the scope of the inventive concept of this article.Therefore, the above-mentioned subject matter should be understood as exemplary and not restrictive, and the example of the embodiment is intended to cover all these modifications, improvements and other embodiments that fall within the spirit and scope of the inventive concept.Therefore, to the maximum extent permitted by law, the scope of the inventive concept should be determined by the widest permissible interpretation of the present disclosure including the examples of the embodiments and their equivalents, and should not be limited to or restricted to the previous specific embodiments.

Claims

1. A computer-implemented method for causally encoding channel state information, performed by a first computing device, the method comprising: Obtaining (1402) an indication of an erroneous reporting of channel state information CSI; retrieving (1404) parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining (1406) a measurement value of the CSI data; activating (1408) a machine learning (ML) model for CSI reporting, the ML model comprising causal relationships between the parameter data and the measured values ​​of the CSI data; Encoding the measurement value of the CSI data (1410) to obtain encoded CSI measurement data; applying (1412) the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; as well as The adapted CSI measurement data is sent (1414) to a second computing device.

2. The method according to claim 1, wherein: The ML model includes a structural causal model SCM.

3. The method according to any one of claims 1 to 2, wherein: The ML model models the CSI measurement data as a function of the plurality of specific parameters and the plurality of general parameters.

4. The method according to claim 3, wherein: The ML model includes a directed acyclic graph including a plurality of nodes corresponding to the CSI measurement data and a plurality of directed edges representing causal parent-child relationships.

5. The method according to any one of claims 1 to 4, wherein: The obtaining (1402) includes at least one of: (i) detecting an error report of the CSI, and (ii) receiving a message including the error report of the CSI from the second computing device.

6. The method according to any one of claims 1 to 5, wherein: The plurality of general parameters include at least one or more of: (i) a speed of the first computing device, (ii) a location of the first computing device, (iii) a moving direction of the first computing device, (iv) line-of-sight or non-line-of-sight propagation of signals between the first computing device and the second computing device, (v) spatial channel correlation, (vi) one or more subcarrier frequencies, (vii) environmental conditions, (viii) available bandwidth, and (ix) a current state of the first computing device.

7. The method according to any one of claims 1 to 6, wherein: The multiple specific parameters include at least one or more of the following: (i) a measurement of perceived interference, (ii) a threshold for a first computing device of a specific vendor, (iii) a threshold for a second computing device of a specific vendor, (iv) available power at the first computing device, (v) a nominal power parameter of the second computing device, (vi) an input power of the second computing device, and (vii) a specified power level of the first computing device.

8. The method according to any one of claims 1 to 7, further comprising: The ML model is constructed (1400) using a graph neural network (GNN), the plurality of specific parameters, the plurality of general parameters, and CSI data, wherein the constructed ML model models how the distribution of the CSI measurement data changes as the specific parameters and the general parameters change.

9. The method according to claim 8, wherein: The construction (1400) includes: Encoding the multiple specific parameters and the multiple general parameters respectively; Inputting to the GNN (i) a plurality of encoded specific parameters as tensors, (ii) a plurality of encoded general parameters as tensors, and (iii) CSI data; and The GNN is trained using an adaptive algorithm and an error metric, wherein the trained GNN captures causal relationships among the multiple specific parameters, the multiple general parameters, and the CSI data.

10. The method according to any one of claims 1 to 9, wherein: The ML model is trained in a federated learning setting.

11. The method according to any one of claims 1 to 10, wherein: The first computing device comprises a user equipment UE, and the second computing device comprises a network node.

12. A computer-implemented method for decoding causally encoded channel state information performed by a second computing device, the method comprising: Receiving (1500) adapted channel state information (CSI) data from a first computing device, wherein the adapted CSI data is output from a machine learning (ML) model for CSI reporting, the ML model including causal relationships between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and measurement values ​​of the CSI data; decoding (1502) the adapted CSI measurement data to obtain decoded CSI; and The decoded CSI is used (1504) for resource allocation.

13. A computer-implemented method for causally encoding and decoding channel state information (CSI), performed by a second computing device, the method comprising: Detect (1602) an erroneous report of channel state information CSI of the first computing device; retrieving (1604) parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtain (1606) a measurement value of CSI data from the first computing device; activating (1608) a machine learning (ML) model for CSI reporting, the ML model including causal relationships between the parameter data and the measured CSI data; Encoding the measured CSI data (1610) to obtain encoded CSI measurement data; as well as The ML model is applied (1612) to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

14. The method according to claim 13, further comprising: Decoding (1614) the adapted CSI measurement data to obtain decoded CSI; as well as The decoded CSI is used (1616) for resource allocation.

15. The method according to any one of claims 13 to 14, wherein: The ML model includes a structural causal model SCM.

16. The method according to any one of claims 13 to 15, wherein: The ML model models the CSI measurement data as a function of the plurality of specific parameters and the plurality of general parameters.

17. The method according to claim 16, wherein: The ML model includes a directed acyclic graph including a plurality of nodes corresponding to the CSI measurement data and a plurality of directed edges representing a causal parent-child relationship between the CSI measurement data and the parameter data.

18. The method according to any one of claims 13 to 17, wherein: The plurality of general parameters include at least one or more of: (i) a speed of the first computing device, (ii) line-of-sight or non-line-of-sight propagation of signals between the first computing device and the second computing device, (iii) spatial channel correlation, (iv) one or more subcarrier frequencies, (v) environmental conditions, (vi) available bandwidth, and (vii) a current state of the first computing device.

19. The method according to any one of claims 13 to 18, wherein: The multiple specific parameters include at least one or more of the following: (i) a measurement of perceived interference, (ii) a threshold for a first computing device of a specific vendor, (iii) a threshold for a second computing device of a specific vendor, (iv) available power at the first computing device, (v) a nominal power parameter of the second computing device, (vi) an input power of the second computing device, and (vii) a specified power level of the first computing device.

20. The method according to any one of claims 13 to 19, further comprising: The ML model is constructed (1600) using a graph neural network (GNN), the plurality of specific parameters, the plurality of general parameters, and CSI data, wherein the constructed ML model models how the distribution of the CSI measurement data changes as the specific parameters and the general parameters change.

21. The method according to claim 20, wherein: The construction (1600) includes: Encoding the multiple specific parameters and the multiple general parameters respectively; Inputting into the GNN (i) a plurality of encoded specific parameters as tensors, (ii) a plurality of encoded general parameters as tensors, and (iii) CSI data; The GNN is trained using an adaptive algorithm and an error metric, wherein the trained GNN captures causal relationships among the multiple specific parameters, the multiple general parameters, and the CSI data.

22. The method according to any one of claims 13 to 21, wherein: The ML model is trained in a federated learning setting.

23. The method according to any one of claims 13 to 22, wherein: The first computing device comprises a user equipment UE, and the second computing device comprises a network node.

24. A first computing device (200, 18300), configured to perform causal encoding on channel state information, the first computing device comprising: Processing circuit (18302); A memory (18304) coupled to the processing circuit, wherein the memory includes instructions that, when executed by the processing circuit, cause the first computing device to perform operations, the operations comprising: Obtaining an indication of an erroneous reporting of channel state information (CSI); retrieving parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining measurement values ​​of CSI data; activating a machine learning (ML) model for CSI reporting, the ML model comprising a causal relationship between the parameter data and the measured values ​​of the CSI data; encoding the measurement value of the CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; and The adapted CSI measurement data is sent to a second computing device.

25. The first computing device of claim 24, wherein: The memory comprises instructions which, when executed by the processing circuit, cause the first computing device to perform further operations, the further operations comprising operations according to any one of claims 2 to 11.

26. A first computing device (200, 18300), configured to causally encode channel state information, the first computing device being adapted to perform operations, the operations comprising: Obtaining an indication of an erroneous reporting of channel state information (CSI); retrieving parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining measurement values ​​of CSI data; activating a machine learning (ML) model for CSI reporting, the ML model comprising a causal relationship between the parameter data and the measured values ​​of the CSI data; encoding the measurement value of the CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; as well as The adapted CSI measurement data is sent to a second computing device.

27. The first computing device of claim 26, adapted to perform the further operations of any one of claims 2 to 11.

28. A computer program comprising program code to be executed by a processing circuit (18302) of a first computing device (200, 18300), the first computing device (200, 18300) being configured to causally encode channel state information, whereby execution of the program code causes the first computing device to perform operations comprising: Obtaining an indication of an erroneous reporting of channel state information (CSI); retrieving parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining measurement values ​​of CSI data; activating a machine learning (ML) model for CSI reporting, the ML model comprising a causal relationship between the parameter data and the measured values ​​of the CSI data; encoding the measurement value of the CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; as well as The adapted CSI measurement data is sent to a second computing device.

29. A computer program according to claim 28, whereby execution of the program code causes the first computing device to perform the operations of any one of claims 2 to 11.

30. A computer program product comprising a non-transitory storage medium (18304), the non-transitory storage medium comprising program code to be executed by processing circuitry (18302) of a first computing device (200, 18300), the first computing device (200, 18300) being configured to causally encode channel state information, whereby execution of the program code causes the first computing device to perform operations comprising: Obtaining an indication of an erroneous reporting of channel state information (CSI); retrieving parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining measurement values ​​of CSI data; activating a machine learning (ML) model for CSI reporting, the ML model comprising a causal relationship between the parameter data and the measured values ​​of the CSI data; encoding the measurement value of the CSI data to obtain encoded CSI measurement data; applying the ML model to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data; as well as The adapted CSI measurement data is sent to a second computing device.

31. The computer program product of claim 30, whereby execution of the program code causes the first computing device to perform the operations of any one of claims 2 to 11.

32. A second computing device (202, 19300), configured to decode causally encoded channel state information, the second computing device comprising: Processing circuit (19302); A memory (19304) coupled to the processing circuit, wherein the memory includes instructions that, when executed by the processing circuit, cause the second computing device to perform operations, the operations comprising: Receiving adapted channel state information (CSI) data from a first computing device, wherein the adapted CSI data is output from a machine learning (ML) model for CSI reporting, the ML model including a causal relationship between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and a measurement value of the CSI data; decoding the adapted CSI measurement data to obtain decoded CSI; and The decoded CSI is used for resource allocation.

33. A second computing device (202, 19300), configured to decode causally encoded channel state information, the second computing device being adapted to perform operations, the operations comprising: Receiving adapted channel state information (CSI) data from a first computing device, wherein the adapted CSI data is output from a machine learning (ML) model for CSI reporting, the ML model including a causal relationship between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and a measurement value of the CSI data; decoding the adapted CSI measurement data to obtain decoded CSI; and The decoded CSI is used for resource allocation.

34. A computer program comprising program code to be executed by processing circuitry (19302) of a second computing device (202, 19300), the second computing device (202, 19300) being configured to decode causally encoded channel state information, whereby execution of the program code causes the second computing device to perform operations comprising: Receiving adapted channel state information (CSI) data from a first computing device, wherein the adapted CSI data is output from a machine learning (ML) model for CSI reporting, the ML model including a causal relationship between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and a measurement value of the CSI data; decoding the adapted CSI measurement data to obtain decoded CSI; and The decoded CSI is used for resource allocation.

35. A computer program product comprising a non-transitory storage medium (19304), the non-transitory storage medium comprising program code to be executed by processing circuitry (19302) of a second computing device (202, 19300), the second computing device (202, 19300) being configured to decode causally encoded channel state information, whereby execution of the program code causes the second computing device to perform operations comprising: Receiving adapted channel state information (CSI) data from a first computing device, wherein the adapted CSI data is output from a machine learning (ML) model for CSI reporting, the ML model including a causal relationship between parameter data, the parameter data including a plurality of specific parameters and a plurality of general parameters and a measurement value of the CSI data; decoding the adapted CSI measurement data to obtain decoded CSI; and The decoded CSI is used for resource allocation.

36. A second computing device (202, 19300), configured to perform causal encoding and decoding on channel state information CSI, the second computing device comprising: Processing circuit (19302); A memory (19304) coupled to the processing circuit, wherein the memory includes instructions that, when executed by the processing circuit, cause the second computing device to perform operations, the operations comprising: Detecting an erroneous report of channel state information CSI of the first computing device; retrieving parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining a measurement value of CSI data from the first computing device; activating a machine learning (ML) model for CSI reporting, the ML model including a causal relationship between the parameter data and the measured CSI data; encoding the measured CSI data to obtain encoded CSI measurement data; and The ML model is applied to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

37. The second computing device of claim 36, wherein: The memory comprises instructions which, when executed by the processing circuit, cause the second computing device to perform further operations, the further operations comprising operations according to any one of claims 14 to 23.

38. A second computing device (202, 19300), configured to perform causal encoding and decoding of channel state information CSI, the second computing device being adapted to perform operations, the operations comprising: Detecting an erroneous report of channel state information CSI of the first computing device; retrieving parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining a measurement value of CSI data from the first computing device; activating a machine learning (ML) model for CSI reporting, the ML model including a causal relationship between the parameter data and the measured CSI data; encoding the measured CSI data to obtain encoded CSI measurement data; as well as The ML model is applied to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

39. The first computing device of claim 38, adapted to perform the further operations of any of claims 14 to 23.

40. A computer program comprising program code to be executed by a processing circuit (19302) of a second computing device (202, 19300), the second computing device (202, 19300) being configured to causally encode and decode channel state information (CSI), whereby execution of the program code causes the second computing device to perform operations comprising: Detecting an erroneous report of channel state information CSI of the first computing device; retrieving parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining a measurement value of CSI data from the first computing device; activating a machine learning (ML) model for CSI reporting, the ML model including a causal relationship between the parameter data and the measured CSI data; encoding the measured CSI data to obtain encoded CSI measurement data; as well as The ML model is applied to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

41. A computer program according to claim 40, whereby execution of the program code causes the second computing device to perform the operations of any one of claims 14 to 23.

42. A computer program product comprising a non-transitory storage medium (19304) comprising program code to be executed by a processing circuit (19302) of a second computing device (202, 19300). 19300), the second computing device (202, 19300) is configured to perform causal encoding and decoding on channel state information CSI, whereby the execution of the program code causes the second computing device to perform operations, the operations comprising: Detecting an erroneous report of channel state information CSI of the first computing device; retrieving parameter data, the parameter data comprising a plurality of specific parameters and a plurality of general parameters; Obtaining a measurement value of CSI data from the first computing device; activating a machine learning (ML) model for CSI reporting, the ML model including a causal relationship between the parameter data and the measured CSI data; encoding the measured CSI data to obtain encoded CSI measurement data; and The ML model is applied to the parameter data and the encoded CSI measurement data to obtain adapted CSI measurement data.

43. The computer program product of claim 42, whereby execution of the program code causes the second computing device to perform the operations of any one of claims 14 to 23.

Images