channel state information

Abstract

Embodiments of the present disclosure relate to channel state information. A method, apparatus and computer program are described comprising: receiving data transmitted on a wireless channel, the data encoding a first estimate of channel state information for the wireless channel; decoding the data to generate a second estimate of the channel state information for the wireless channel; and generating a channel state information output by modifying the second estimate of the channel state information based at least in part on a first error indication.

Description

Technical Field

[0001] This specification relates to channel state information in wireless communication systems. Background Technology

[0002] This specification relates to channel state information in wireless communication systems. For example, an estimate of channel state information may be transmitted by a device (e.g., a user equipment) on a channel of the wireless communication system. Summary of the Invention

[0003] In a first aspect, this specification describes an apparatus comprising: components for receiving data transmitted by a device of a wireless communication system on a channel of the wireless communication system (the device may be a user equipment or user gear of a mobile communication system), wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; components for decoding the data to generate a second estimate of the channel state information of the wireless channel; and components for generating a channel state information output by modifying the second estimate of the channel state information at least in part based on a first error indication (e.g., a binary scalar), the first error indication indicating an estimation error in the first channel state information estimate generated at the device.

[0004] In a second aspect, this specification describes an apparatus comprising: components for receiving data transmitted by a device of a wireless communication system on a channel of the wireless communication system (the device may be a user equipment or user gear of a mobile communication system), wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; components for decoding the data to generate a second estimate of the channel state information of the wireless channel; and components for generating a channel state information output by modifying the second estimate of the channel state information at least in part based on partially correcting for errors introduced in the data transmission over the wireless connection between the device and the network node. Alternatively or additionally, the second estimate of the channel state information may be modified at least in part based on a first error indication (e.g., a binary scalar) indicating an estimation error in the first channel state information estimate generated at the device.

[0005] Some example embodiments of the first and second aspects also include components for generating a first error indication. The components for generating the first error indication may include a first machine learning module (e.g., another neural network). Some example embodiments also include components for training the first machine learning module by minimizing a loss function comprising labeled data.

[0006] In some example embodiments of the first and second aspects, the components for generating the channel state information output include a second machine learning module (e.g., a second neural network). Some example embodiments of the first and second aspects also include components for training the second machine learning module based on a comparison of the generated channel state information with labeled channel state information data.

[0007] In some example embodiments of the first and second aspects, the components for generating the channel state information output are configured to at least partially correct for errors introduced in the data transmitted over the wireless connection between the device and the network node. The apparatus may also include components for detecting errors introduced in the data transmitted over the wireless connection between the device and the network node. This detection component may be implemented using a machine learning module, a neural network, or some other module. For example, the second machine learning module discussed above may be used.

[0008] In the first and second aspects, the component may include: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the execution of the device.

[0009] In a third aspect, this specification describes a method comprising: receiving data transmitted by a device of a wireless communication system on a channel of the wireless communication system (the device may be a user equipment or user gear of a mobile communication system), wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; decoding the data to generate a second estimate of the channel state information of the wireless channel; and generating a channel state information output by modifying the second estimate of the channel state information at least in part based on a first error indication (e.g., a binary scalar), the first error indication indicating an estimation error in the first channel state information estimate generated at the device. The method may further include generating the first error indication. Furthermore, generating the channel state information output may include at least partially correcting for errors introduced in the data over the wireless connection between the device and the network node.

[0010] In a fourth aspect, this specification describes a method comprising: receiving data transmitted by a device of a wireless communication system on a channel of the wireless communication system (the device may be a user equipment or user gear of a mobile communication system), wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; decoding the data to generate a second estimate of the channel state information of the wireless channel; and generating a channel state information output by modifying the second estimate of the channel state information at least in part based on partially correcting for errors introduced in the data transmission over the wireless connection between the device and the network node. Alternatively or additionally, the second estimate of the channel state information may be modified at least in part based on a first error indication (e.g., a binary scalar) indicating an estimation error in the first channel state information estimate generated at the device. The method may further include generating the first error indication. Furthermore, generating the channel state information output may include at least partially correcting for errors introduced in the data over the wireless connection between the device and the network node.

[0011] A first machine learning module (e.g., another neural network) can be used to generate a first error indication. Some example embodiments of the third and fourth aspects further include training the first machine learning module by minimizing a loss function that includes labeled code data.

[0012] A second machine learning module (e.g., a second neural network) can be used to generate channel state information output. Some example embodiments of the third and fourth aspects further include training the second machine learning module based on a comparison of the generated channel state information with labeled channel state information data.

[0013] Some example embodiments of the third and fourth aspects also include detecting errors introduced into the data transmitted over the wireless connection between the device and the network node. The detection component can be implemented using a machine learning module, a neural network, or some other module. For example, the second machine learning module discussed above can be used.

[0014] In a fifth aspect, this specification describes computer-readable instructions that, when executed by a computing device, cause the computing device to perform at least any of the methods described herein (including the methods described in the third or fourth aspect above).

[0015] In a sixth aspect, this specification describes a computer-readable medium (such as a non-transient computer-readable medium) comprising program instructions stored thereon for performing (at least) any of the methods described herein (including the methods of the third or fourth aspects above).

[0016] In a seventh aspect, this specification describes an apparatus comprising: at least one processor; and at least one memory including computer program code that, when executed by the at least one processor, causes the apparatus to perform (at least) any of the methods described herein (including the methods of the third or fourth aspects above).

[0017] In an eighth aspect, this specification describes an apparatus comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus to perform at least any of the methods described herein (including the methods of the third or fourth aspects described above).

[0018] In a ninth aspect, this specification describes a computer program including instructions that, when executed by an apparatus, cause the apparatus to: receive data transmitted by a device of a wireless communication system on a channel of the wireless communication system (the device may be a user equipment or user gear of a mobile communication system), wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; decode the data to generate a second estimate of the channel state information of the wireless channel; and generate a channel state information output by modifying the second estimate of the channel state information at least in part based on a first error indication (e.g., a binary scalar), the first error indication indicating an estimation error in the first channel state information estimate generated at the device. The apparatus is also caused to generate the first error indication. Furthermore, generating the channel state information output may include at least partially correcting for errors introduced in the data over the wireless connection between the device and the network node.

[0019] In a tenth aspect, this specification describes a computer program comprising instructions that, when executed by an apparatus, cause the apparatus to: receive data transmitted by a device of a wireless communication system on a channel of the wireless communication system (the device may be a user equipment or user gear of a mobile communication system), wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; decode the data to generate a second estimate of the channel state information of the wireless channel; and generate a channel state information output by modifying the second estimate of the channel state information at least in part based on partially correcting errors introduced in the data transmission over the wireless connection between the device and the network node. Alternatively or additionally, the second estimate of the channel state information may be modified at least in part based on a first error indication (e.g., a binary scalar) indicating an estimation error in the first channel state information estimate generated at the device. The apparatus is also caused to generate the first error indication. Furthermore, generating the channel state information output may include at least partially correcting errors introduced in the data over the wireless connection between the device and the network node.

[0020] In the eleventh aspect, this specification describes: a receiver module (or some other component) for receiving data transmitted by a device of a wireless communication system on a channel of the wireless communication system (the device may be a user equipment or user gear of a mobile communication system), wherein the data encodes a first estimate of channel state information of a wireless channel determined at the device; a decoder, such as an AIML decoder (or some other component), for decoding the data to generate a second estimate of the channel state information of the wireless channel; and a processor, such as one or more machine learning modules (or some other component), for generating a channel state information output by modifying the second estimate of the channel state information at least in part based on a first error indication (e.g., a binary scalar), the first error indication indicating an estimation error in the first channel state information estimate generated at the device.

[0021] In a twelfth aspect, this specification describes: a receiver module (or some other component) for receiving data transmitted by a device of a wireless communication system on a channel of the wireless communication system (the device may be a user equipment or user gear of a mobile communication system), wherein the data encodes a first estimate of channel state information of a wireless channel determined at the device; a decoder, such as an AIML decoder (or some other means), for decoding the data to generate a second estimate of the channel state information of the wireless channel; and a processor, such as one or more machine learning modules (or some other means), for generating a channel state information output by modifying the second estimate of the channel state information by at least partially based on partially correcting errors introduced in the data transmission over the wireless connection between the device and the network node. Attached Figure Description

[0022] Exemplary embodiments will now be described by way of non-limiting examples with reference to the following schematic diagrams, in which:

[0023] Figures 1 to 4 This is a block diagram illustrating a system according to an example embodiment;

[0024] Figures 5 to 8 This is a flowchart illustrating an algorithm according to an example embodiment;

[0025] Figure 9 This is a block diagram illustrating a system according to an example embodiment;

[0026] Figure 10 These are schematic diagrams of components from one or more example embodiments previously described; and

[0027] Figure 11 A tangible medium for storing computer-readable code is shown, which, when run by a computer, can execute methods according to the example embodiments described herein. Detailed Implementation

[0028] The scope of protection sought by the various embodiments of the present invention is stated in the independent claims. Embodiments and features described in this specification that do not fall within the scope of the independent claims (if any) are to be interpreted as examples useful for understanding the various embodiments of the invention.

[0029] In the description and accompanying drawings, the same reference numerals always denote the same elements.

[0030] Figure 1This is a block diagram illustrating a system generally indicated by reference numeral 10 according to an exemplary embodiment. The system includes a device (such as a user equipment (UE)) 12, a channel 14, and a network node 16. The device 12 and the network node 16 communicate bidirectionally with each other via the channel 14.

[0031] In systems such as System 10, Channel State Information (CSI) may be needed for a variety of purposes, such as accurate precoding in massive MIMO communications using Frequency Division Duplex (FDD) schemes. For example, accurate CSI can be used to enable base stations (BSs) to achieve higher signal-to-noise ratios (SNR) and improve channel capacity. However, in many implementations of System 10 (e.g., an FDD network), only device 12 is able to estimate the downlink CSI. In such systems, the estimated CSI is typically shared with network node 16 via channel 14, which introduces communication overhead into the system.

[0032] To reduce this overhead, CSI data can be compressed and / or quantized before transmission. For example, CSI data can be compressed (e.g., using vector quantization) to generate a codebook that achieves a high compression ratio (CR) (i.e., the ratio of compressed size to uncompressed size). (The compression ratio can be a scalar in the range (0, 1], where a lower CR indicates more compression.) Alternatively or additionally, CSI data can be quantized such that, for example, device 12 transmits the quantized bits to network node 16 via channel 14, network node 16 including a dequantizer to generate a signal that a decoder can use to reconstruct the original channel state information estimate.

[0033] Schemes used to reduce transmission overhead (such as data compression and / or data quantization) may limit the accuracy of CSI reconstruction.

[0034] Figure 2 This is a block diagram illustrating a system generally indicated by reference numeral 20 according to an exemplary embodiment. System 20 includes a channel estimator 22, an encoder 24, and a transmitter 26, which form part of an exemplary implementation of the device 12 described above. The channel estimator 22 can be used to estimate downlink channel state information (CSI). An encoded version of the estimated CSI is transmitted using transmitter 26 (e.g., from device 12 to network node 16 on channel 14).

[0035] Figure 3 This is a block diagram illustrating a system generally indicated by reference numeral 30 according to an exemplary embodiment. System 30 includes a receiver 32 and a decoder 34, which may form part of the exemplary embodiment of network node 16 described above. Receiver 32 may receive encoded CSI data transmitted by system 20. Decoder 34 decodes the encoded transmitted CSI to reconstruct the CSI estimated by channel estimator 22.

[0036] Encoder 24 and decoder 34 can work together as an autoencoder (AE), where encoder 24 is used to generate a compressed representation of the input CSI and decoder 34 is used to reconstruct the CSI from the compressed information.

[0037] The encoder and decoder for CSI compression based on autoencoders can be implemented using artificial intelligence or machine learning models (AIML). These models can be trained in various ways, such as:

[0038] Joint training of the dual-ended autoencoder model at one side / entity (e.g., at the UE / device side or network side of System 10). This can be referred to as Type 1 training.

[0039] Joint training of the dual-ended autoencoder model on the network side and the UE / device side of System 10, respectively. This can be referred to as Type 2 training.

[0040] Separate training is performed on the network side and the UE / device side. The UE / device side CSI generation part is trained separately, and the network side CSI reconstruction part is trained separately. This can be referred to as Type 3 training.

[0041] Joint training can refer to the code generation model (encoder) and the reconstruction model (decoder) being trained in the same loop used for forward and backward propagation. Joint training can be performed at a single node or across multiple nodes (e.g., through gradient exchange between nodes).

[0042] Individual training can include sequential training that begins on the UE side, or sequential training that begins on the network side at the UE.

[0043] In system 20, the output of channel estimator 22 provides an imperfect channel estimate represented by eCSI, as follows:

[0044] eCSI=CSI+∈1

[0045] Here, ∈1 represents the error term introduced in imperfect downlink channel estimation. Note that ∈1 is typically the primary source of error in CSI feedback.

[0046] Then, the eCSI is encoded into a bit vector C by encoder 24 (which can form part of the AIML autoencoder).

[0047] Code C is prepared for transmission by transmitter module 26. For example, code C can undergo OFDM transmission on the uplink control channel (e.g., the bits can be encoded using Polar codes, modulated onto QAM symbols, and then modulated into an OFDM waveform).

[0048] The transmitted waveform (e.g., OFDM waveform) travels over a wireless propagation channel (e.g., channel 14) to receiver 32 of system 30. During transmission, the waveform is contaminated by noise and interference. At receiver 32, the signal undergoes OFDM reception (using receiver 32) and decoding (using decoder 34) to generate an estimated code. The estimation quality depends on the wireless conditions and the robustness of the transmission scheme.

[0049]

[0050] Here, ∈2 is the error at the input of decoder 34. In some example embodiments, ∈2 may be a rare event due to the very high error detection / correction probability of conventional channel coding with CRC check. However, depending on the signal-to-noise ratio (SNR) of the received signal and the code length and code rate considered for conventional channel coding, ∈2 may be higher. If ∈2 is available, and we successfully correct the received code, then there is no need to schedule retransmission, which reduces latency and saves resources.

[0051] The estimated code is decoded by decoder 34, and the reconstructed eCSI (represented as) is output. ).therefore, It is an imperfect estimate of the actual CSI, and the imperfections are mainly due to two reasons:

[0052] The error ∈ 1 is due to the imperfect downlink channel estimation in the UE (i.e., the first block in which eCSI is generated).

[0053] Errors based on UL caused by imperfect uplink OFDM reception (which introduces errors in the estimated code in the gNB).

[0054] Imperfect Reconstruction It can be used, for example, to select a pre-encoder. Wherein, depends on... Ultimately, however inaccurate the choice may be, it may become suboptimal.

[0055] In order to provide an accurate CSI feedback architecture that is relatively unaffected by the severity of channel conditions (both DL and UL), it is expected that network node 16 will reconstruct an accurate version of the CSI.

[0056] Figure 4 This is a block diagram illustrating a system generally indicated by reference numeral 40 according to an example embodiment.

[0057] The system includes a channel estimator 41, an encoder 42, an OFDM transmitter module 43, a wireless channel 44, an OFDM receiver module 45, a decoder 46, an error detection module 47, and an error correction module 48.

[0058] The channel estimator 41, encoder 42, and transmitter module 43 are similar to the channel estimator 22, encoder 24, and transmitter 26, and may form part of device 12 in the system 10 described above. The receiver module 45, decoder 46, error detection module 47, and error correction module 48 may form part of network node 16 in the system 10 described above.

[0059] OFDM receiver 45 implements components for receiving data transmitted by a device (e.g., device 12) on channel 44 of the wireless communication system. This data is encoded with a first estimate of channel state information of the wireless channel determined at the device (i.e., as determined at channel estimator 41).

[0060] As described above with reference to system 20, the channel estimator 41 is an imperfect channel estimator that provides the following imperfect channel estimate, denoted by eCSI:

[0061] eCSI=CSI+∈1

[0062] Here, ∈1 is the error term introduced by imperfect downlink channel estimation.

[0063] Then, encoder 42 encodes the eCSI into a bit vector C.

[0064] Code C is prepared for transmission by OFDM transmitter module 43. At OFDM receiver 45, the following estimated code is generated. The estimation quality depends on the robustness of the wireless conditions and the transmission scheme:

[0065]

[0066] Estimated code It is passed through decoder 46 (which may be the decoder part of an AIML autoencoder with encoder 42) to generate the reconstructed CSI estimate. Then, the reconstructed CSI estimate The error is sent to error correction module 48, which is designed to remove some or all of the uplink and downlink errors from the reconstructed code.

[0067] Refactoring code It is also routed to error detection module 47, which detects errors in the reconstructed code and outputs a scalar e∈[0,1], where if Then e = 1.

[0068] Error correction module 48 receives:

[0069] Reconstructed code output by OFDM receiver 45

[0070] Reconstruction output from decoder 46 as well as

[0071] The detected error e is output by the error detection module 47.

[0072] Error correction module 48 attempts to remove errors in the reconstructed CSI estimate. The uplink and downlink errors in the data are used to generate an estimate that closely approximates the true CSI.

[0073] Therefore, decoder 46 generates a second estimate of the channel state information of the wireless channel. Furthermore, the error correction module generates channel state information output by modifying a second estimate of the channel state information based at least in part on the first error indication (e). The first error indication may indicate the estimation error (i.e., downlink error) in the first channel state information estimation generated at the device. Alternatively or additionally, the first error indication may indicate the uplink error introduced in the data transmitted over the wireless connection between the device and the network node.

[0074] It should be noted that although the decoder 46, error detection module 47, and error correction module 48 are shown as three separate modules, two or more of these modules can be combined into a single module (e.g., into a single neural network). For example, error detection module 47 and error correction module 48 can be integrated. However, separating these modules may have advantages; at the very least, the training of the modules can be more easily separated.

[0075] As discussed in detail below, the error correction module 48 can be a neural network (NN) or a machine learning module that can be trained to distinguish the following:

[0076] (At channel estimation module 41) DL channel estimation error; and

[0077] UL receiving error.

[0078] The error detection module 47 (for generating the first error indication) can also be implemented using a neural network (NN) or machine learning module.

[0079] Figure 5This is a flowchart illustrating an algorithm, generally indicated by reference numeral 50, according to an example embodiment. Algorithm 50 can be implemented on the device side of system 40 (i.e., using channel estimator 41, encoder 42, and OFDM transmitter module 43).

[0080] Algorithm 50 begins with operation 52, in which a channel estimate (e.g., an imperfect channel estimate) is obtained. The channel estimate can be generated using conventional methods or machine learning methods.

[0081] In operation 54, the channel estimate is encoded (e.g., using the encoder portion of an autoencoder). The encoder can be a machine learning module and can be trainable.

[0082] In operation 56, the encoded channel estimate is transmitted using OFDM or some other transmission arrangement. For example, the output of operation 54 can be encoded, modulated onto QAM symbols, and subsequently transmitted as an OFDM waveform.

[0083] Figure 6 This is a flowchart illustrating an algorithm, generally indicated by reference numeral 60, according to an exemplary embodiment. Algorithm 60 can be implemented on the network side of system 40 (e.g., using OFDM receiver module 45, decoder 46, error detection module 47, and error correction module 48).

[0084] Algorithm 60 begins at operation 62, in which the signal transmitted in operation 56 is received (e.g., using OFDM or some other receiving technique).

[0085] In operation 64, the channel estimate (as initially generated in operation 52) is detected and corrected (e.g., using one or more neural networks or machine learning algorithms). As discussed in detail above, the correction can correct for errors introduced in operation 52 by imperfect channel estimation, and / or errors introduced during the encoding, transmission, and decoding of the channel estimate. Detection and / or correction blocks based on trainable machine learning can be used to detect errors in the received code and / or generate modified channel estimates.

[0086] Finally, in operation 66, the channel estimate obtained in operation 64 is used in some way. For example, the channel estimate can be used to select a precoder for massively multi-input multiple-output (MIMO) communication.

[0087] In system 40, the error detection module 47 is typically a neural network that outputs a binary scalar indicating the error. The error detection module 47 can be trained for a binary classification task.

[0088] Binary cross-entropy (CE) can be used as a cost function to train the binary classifier implemented as error detection module 47. The labels used for training can be obtained by sharing the code C from a device (e.g., device 12) to a network node (e.g., node 16) using a more reliable link or by considering multiple transmissions or sharing in the cloud. The label can then be written as...

[0089]

[0090] The output e of the ML block and the label e L The cross-entropy (CE) loss function between them is

[0091] L CE =-(e (l) loge+(1-e (l) log(1-e))

[0092] To train the error correction module 48, the mean square error (MSE) or squared generalized cosine similarity (SGCS) between the CSI and the optimal / expected CSI can be reconstructed. Note that the optimal / expected CSI used as the label can be obtained by a simulator or by considering a more complex channel estimator compared to the channel estimator used for inference in the user equipment. For example, the user equipment uses least squares (LS) as the channel estimator in inference. However, the user equipment uses both LS and LMMSE channel estimators for data collection during the training phase.

[0093] Figure 7 This is a flowchart illustrating an algorithm, generally indicated by reference numeral 70, according to an example embodiment. Algorithm 70 can be used to train error detection module 47.

[0094] Algorithm 70 begins with operation 72, in which a code related to channel estimation is received (e.g., a reconstructed code output by OFDM receiver 45). ).

[0095] In operation 74 of algorithm 70, the transmitted code (e.g., a more reliable version of code C) is obtained. The code received in this operation represents labeled data that can be used for training.

[0096] At operation 76 of algorithm 70, error detection module 47 is trained by minimizing the loss function based on the codes received in operations 72 and 74.

[0097] Of course, Algorithm 70 is highly illustrative and can be modified in many ways. For example, operations 72 and 74 can be reversed, or they can be implemented in parallel.

[0098] Figure 8This is a flowchart illustrating an algorithm, generally indicated by reference numeral 80, according to an example embodiment. Algorithm 80 can be used to train error correction module 48.

[0099] Algorithm 80 begins with operation 82, in which (e.g., by an instance of error correction module 48) a channel state information estimate is generated.

[0100] In operation 84 of algorithm 80, tagged channel state information data is received. The tagged data is considered to be more accurate (or more reliable) than the data obtained in operation 82.

[0101] In operation 86 of algorithm 80, the error correction module 48 is trained by minimizing the loss function based on the channel state information obtained in operations 82 and 84.

[0102] Of course, Algorithm 80 is highly illustrative and can be modified in many ways. For example, operations 82 and 84 can be reversed, or they can be implemented in parallel.

[0103] Figure 9 This is a block diagram illustrating a system generally indicated by reference numeral 90 according to an exemplary embodiment. System 90 has many similarities to system 40 described above.

[0104] In system 90, the error detection module 97 of system 90 provides two sets of outputs. These outputs include:

[0105] Scalar e∈[0,1] (as described above), where if Then e = 1 (a smaller value of e means) (The probability is higher);

[0106] binary vector i e This is used to determine which elements / segments in the code are affected by UL channel noise.

[0107] The information in e can help find the code. The index of elements affected by UL channel noise (for scalar quantization) or the index of segments (for vector quantization). In the code If an error is detected in one of the segments / elements, the code correction block can be used to correct the channel code and generate a corrected code.

[0108] Therefore, the error detection module 97 not only detects errors in the reconstructed code, but also outputs the indexes of the segments / elements that need to be corrected.

[0109] Depending on the type of quantization considered (scalar quantization or vector quantization), we can determine the number of elements / segments in the code, which also defines the dimension of the binary vector. The probability of training error detection block 97 indicates that the code generated by the encoder occupies only a portion of the code space. This property has been seen and used in training schemes that first train NWs, where a single decoder decompresses codes from different UEs with different encoders.

[0110] The NN1 block implementing error detection block 97 can be viewed as a multi-class classifier, where the decision of the state for each element / segment is a separate class, resulting in the number of classes equal to the binary vector i. e The dimension of the variable is given. We can use the cross-entropy loss function to train the classifier. For example, the j-th entry of the label can be obtained as...

[0111]

[0112] Different correction mechanisms / algorithms can be considered to correct elements / segments affected by UL channel noise. As an example, the code correction block (e.g., code correction block 48) in the above embodiment can search across all or some possible quantization levels / codewords for the affected segment. The search aims to find quantization levels / codewords that result in new codes that follow the distribution of codes in the training dataset.

[0113] Another possibility is to choose the following quantization level / codeword for the affected segment: if we feed the candidate codeword to the trained error detection module, this quantization level / codeword results in the highest value of e.

[0114] In another example embodiment, a channel correlation function may be additionally fed to the error correction module, and the channel correlation function is one of the following:

[0115] Estimated by gNB and / or predicted using any other UL data services.

[0116] Estimated and reported by the UE, for example, the UE may report maximum Doppler shift and maximum delay spread, and the gNB may use a selected model to determine channel correlation in the time / frequency domain.

[0117] In another embodiment, the error correction module can be located on the device side. For example, the error correction module can be located between the channel estimation module 41 and the encoder module 42. In this case, the error correction module can receive eCSI and optional channel correlation function and estimated SINR as input.

[0118] To implement the different architectures described herein, coordination can be performed between the device side and the network side. For example, if the tasks of CSI correction and bit error detection / correction are distributed between the device side and the network side, the network side should coordinate these tasks, for example, by activating / deactivating UE-based modules. Conversely, if the network decides to perform all the described tasks itself, it can trigger the device to operate in an energy-saving mode (EEM), where, for example:

[0119] The device can use a simple downlink channel estimator, such as least squares (relying on gNB for CSI correction).

[0120] The equipment can postpone the estimation and reporting of channel correlation functions and / or SINR conditions, etc.

[0121] For the sake of completeness, Figure 10 The above diagram illustrates components of one or more example embodiments described above, which will be collectively referred to below as processing system 300. Processing system 300 may be, for example, the apparatus mentioned in the claims.

[0122] The processing system 300 may have a processor 302, a memory 304 tightly coupled to the processor and comprising RAM 314 and ROM 312, and optional user inputs 310 and a display 318. The processing system 300 may include one or more network / device interfaces 308 for connecting to a network / device, such as a wired or wireless modem. The network / device interface 308 may also operate to connect to other devices, such as devices that are not network-side devices. Thus, direct connection between devices is possible without network involvement.

[0123] The processor 302 is connected to each of the other components in order to control their operation.

[0124] Memory 304 may include non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD). Among other things, the ROM 312 of memory 304 stores an operating system 315 and may also store software applications 316. Processor 302 uses RAM 314 of memory 304 to temporarily store data. Operating system 315 may contain code that, when executed by the processor, implements aspects of the algorithms and sequences 50, 60, 70, and 80 described above. Note that in the case of small devices, the memory may be best suited for small size use, i.e., a hard disk drive (HDD) or solid-state drive (SSD) is not always used.

[0125] The processor 302 can take any suitable form. For example, it can be a microcontroller, multiple microcontrollers, a processor, or multiple processors.

[0126] The processing system 300 can be a standalone computer, server, console, or its network. The processing system 300 and the necessary structural components can all be internal to a device, such as an IoT device, i.e., embedded in a very small size.

[0127] In some example embodiments, the processing system 300 may also be associated with external software applications. These applications may be applications stored on a remote server device / device and may run partially or exclusively on the remote server device / device. These applications may be referred to as cloud-hosted applications. The processing system 300 may communicate with the remote server device / device to utilize the software applications stored there.

[0128] Figure 11 A tangible medium for storing computer-readable code in the form of a removable memory unit 365 is illustrated, which, when run by a computer, can execute the methods according to the example embodiments described above. The removable memory unit 365 may be a memory stick, such as a USB memory stick, having an internal memory 366 for storing computer-readable code. The computer system can access the internal memory 366 via a connector 367. Of course, other forms of tangible storage media can be used, as will be apparent to those skilled in the art. The tangible medium can be any device / apparatus capable of storing data / information that can be exchanged between devices / apparatus / networks.

[0129] Embodiments of the present invention may be implemented in software, hardware, application logic, or a combination of software, hardware, and application logic. The software, application logic, and / or hardware may reside in memory or any computer medium. In example embodiments, the application logic, software, or instruction set is maintained on any of a variety of conventional computer-readable media. In the context of this document, "memory" or "computer-readable medium" can be any non-transitory medium or apparatus that may contain, store, communicate, propagate, or transmit instructions used by or in connection with an instruction execution system, apparatus, or device (such as a computer).

[0130] In the relevant context, references to "computer-readable medium," "computer program product," "tangible computer program," or "processor" or "processing circuitry" should be understood to include not only computers with different architectures (e.g., single / multiprocessor architectures and sequential / parallel architectures) but also special-purpose circuits (e.g., field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), signal processing devices / apparatus, and other devices / apparatus). References to computer programs, instructions, code, etc., should be understood to express software (e.g., programmable content of hardware devices / apparatus) used for programmable processor firmware as instructions for a processor or as configured settings or configuration settings for fixed-function devices / apparatus, gate arrays, programmable logic devices / apparatus, etc.

[0131] If necessary, the different functions discussed herein can be executed in different orders and / or simultaneously with each other. Furthermore, one or more of the aforementioned functions can be optional or can be combined, if required. Similarly, it should be understood that... Figures 5 to 8 The flowchart is for illustrative purposes only, and the different operations depicted therein may be omitted, reordered, and / or combined.

[0132] It should be understood that the above-described exemplary embodiments are merely illustrative and do not limit the scope of the invention. Other variations and modifications will be apparent to those skilled in the art after reading this specification.

[0133] Furthermore, the disclosure of this application should be understood to include any novel feature or any novel combination of features disclosed herein, expressly or implicitly, or in any general sense thereof, and during the proceedings of this application or any application derived therefrom, new claims may be formulated to cover any such feature and / or combination of such features.

[0134] While various aspects of the invention are set forth in the independent claims, other aspects of the invention include other combinations of features from the described exemplary embodiments and / or dependent claims with features of the independent claims, and not only those combinations expressly stated in the claims.

[0135] It should also be noted in this document that while various examples have been described above, these descriptions should not be considered limiting. Rather, several changes and modifications may be made without departing from the scope of the invention as defined in the appended claims.

Claims

1. An apparatus for communication, comprising: means for receiving data transmitted by a device of a wireless communication system on a channel of the wireless communication system, wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; means for decoding the data to generate a second estimate of the channel state information of the wireless channel; and means for generating a channel state information output by modifying the second estimate of the channel state information based at least in part on a first error indication that indicates an error in the first estimate of the channel state information generated at the device and an uplink error introduced in the data transmitted on the wireless channel, wherein the means for generating the channel state information output is configured to at least partially correct the uplink error introduced in the data transmitted on the wireless channel based on an identification of elements or segments of an estimated code corresponding to the data that are affected by uplink noise.

2. The apparatus of claim 1, further comprising means for generating the first error indication.

3. The apparatus of claim 2, wherein the means for generating the first error indication comprises a first machine learning module.

4. The apparatus of claim 3, further comprising means for training the first machine learning module by minimizing a loss function that includes labeled code data.

5. The apparatus of claim 1, wherein the first error indication is a binary scalar.

6. The apparatus of claim 1, wherein the means for generating the channel state information output comprises a second machine learning module.

7. The apparatus of claim 6, further comprising means for training the second machine learning module based on a comparison of generated channel state information to labeled channel state information data.

8. The apparatus of claim 1, further comprising means for detecting the uplink error introduced in the data transmitted on the wireless channel between the device and the apparatus.

9. The apparatus of claim 1, wherein the device is a user equipment or user apparatus of a mobile communication system.

10. The apparatus of any one of claims 1-9, wherein the means comprise at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the performance of the apparatus.

11. A method for communication, comprising: receiving data transmitted by a device of a wireless communication system on a channel of the wireless communication system, wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; decoding the data to generate a second estimate of the channel state information of the wireless channel; and generating a channel state information output by modifying the second estimate of the channel state information based at least in part on a first error indication, the first error indication indicating an error in the first estimate of the channel state information generated at the device and an uplink error introduced in the data transmitted over the wireless channel, wherein generating the channel state information output comprises at least partially correcting the uplink error introduced in the data transmitted over the wireless channel based on an identification of elements or segments of an estimated code corresponding to the data that are affected by uplink noise.

12. The method of claim 11, further comprising generating the first error indication.

13. A computer program comprising instructions which, when executed by an apparatus, cause the apparatus to: receive data transmitted over a channel of a wireless communication system by a device of the wireless communication system, wherein the data encodes a first estimate of channel state information of the wireless channel determined at the device; decode the data to generate a second estimate of the channel state information of the wireless channel; and generate a channel state information output by modifying the second estimate of the channel state information based at least in part on a first error indication, the first error indication indicating an error in the first estimate of the channel state information generated at the device and an uplink error introduced in the data transmitted over the wireless channel, wherein generating the channel state information output comprises at least partially correcting the uplink error introduced in the data transmitted over the wireless channel based on an identification of elements or segments of an estimated code corresponding to the data that are affected by uplink noise. ​

Images