Generating demodulation reference signal patterns on resource grid based on channel measurements

Abstract

In particular, a method is disclosed comprising obtaining channel measurements of a radio channel between a first device and a second device measured on a resource grid, generating a demodulation reference signal pattern on the resource grid based on the channel measurements, and transmitting the generated demodulation reference signal pattern to the second device. Computer programs and systems according to the first and second devices are also disclosed.

Description

Technical Field

[0001] The following disclosure relates to the field of wireless networks, or more specifically to systems, apparatuses, and methods for obtaining demodulation reference signal patterns. Background Art

[0002] In a 5G system, a demodulation reference signal (DMRS) can be used for radio channel estimation for demodulation of an associated physical channel or control channel. Current standards define traditional DRMS patterns with mapping types A and B, where mapping type A can only start at symbol 2 or 3 within a time slot, and mapping type B always starts at the first symbol of the scheduling start and length indicator (SLIV). DMRS transmission can be configured through high-layer signaling (RRC setup / reconfiguration), and once configured for a UE, the DMRS pattern is fixed throughout the life cycle of the RRC connection. Summary of the Invention

[0003] The exemplary aspects and exemplary embodiments provided below can achieve enhanced signaling, efficient resource reservation, and interoperability. The generated or derived DMRS pattern can allow for reduced overhead, such as for signaling patterns and / or to achieve a trade-off between DMRS overhead and channel recovery mean square error (MSE). The exemplary aspects and exemplary embodiments provided below can, for example, achieve adaptive scheduling of the DMRS pattern based on the bit error rate (BER) / MSE performance requirements of the UE.

[0004] According to a first exemplary aspect, a first apparatus is disclosed. The first apparatus includes:

[0005] - means for obtaining channel measurements (e.g., at least one channel measurement) of a radio channel between the first apparatus and a second apparatus measured on a resource grid;

[0006] - means for generating (or deriving) a demodulation reference signal pattern on the resource grid based on the channel measurements; and

[0007] - means for sending the generated demodulation reference signal pattern to the second apparatus.

[0008] The first apparatus can, for example, be or include a mobile entity (e.g., a mobile telecommunications device or a mobile phone or a user equipment or a terminal device).

[0009] According to a second exemplary aspect, a second apparatus is disclosed. The second apparatus includes:

[0010] - means for receiving, on a resource grid, a demodulation reference signal pattern from the first apparatus, where the demodulation reference signal pattern is generated based on channel measurements of a radio channel between the first apparatus and the second apparatus measured on the resource grid.

[0011] The second device may for example be or include a server, a server cloud, or a radio access network (RAN) node or a part thereof.

[0012] The device according to any aspect may include: at least one processor; and at least one memory storing instructions which, when executed by the at least one processor, cause the device to perform at least the specified steps.

[0013] The components of the device (e.g., the first and / or second device) may be implemented in hardware and / or software. They may include for example at least one processor for executing processor instructions for performing the required functions, at least one memory storing instructions, or both. Alternatively, they may include for example circuitry designed or configured to implement the required functions, such as implemented in a chipset or a chip, such as an integrated circuit. Generally, the components may include for example one or more processing components or processors.

[0014] As used in this application, the term "circuitry" may refer to one or more or all of the following:

[0015] (a) only hardware circuit implementations (such as for example implementations in analog and / or digital circuitry) and

[0016] (b) combinations of hardware circuits and software, such as (where applicable):

[0017] (i) combinations of (multiple) analog and / or digital hardware circuits with software / firmware, and

[0018] (ii) any part of (multiple) hardware processors with software (including (multiple) digital signal processors), software, and (multiple) memories, which work together to cause a device such as a mobile phone or a server to perform various functions), and

[0019] (c) (multiple) hardware circuits and / or (multiple) processors (such as (multiple) microprocessors or parts of (multiple) microprocessors) that require software (e.g., firmware) to operate, where the software may be absent when not required to operate.

[0020] The definition of circuitry applies to all uses of the term in this application, including in any claims. As another example, as used in this application, the term circuitry also encompasses implementations of only hardware circuits or processors (or multiple processors) or parts of hardware circuits or processors and their (or their) accompanying software and / or firmware. The term circuitry also encompasses, for example and if applicable to a particular claim element, a baseband integrated circuit or a processor integrated circuit for a mobile device or a similar integrated circuit in a server, a cellular network device, or other computing or network device.

[0021] According to a third exemplary aspect, a first method is disclosed. The first method includes:

[0022] - Obtaining channel measurements of a radio channel between a first device and a second device measured on a resource grid;

[0023] - Generating (or deriving) a demodulation reference signal pattern on the resource grid based on the channel measurements; and

[0024] - Transmitting (e.g., to the second device) the generated demodulation reference signal pattern.

[0025] The first method may include steps in which the device of the first exemplary aspect is configured to perform or has components for performing. The method may for example be performed and / or controlled by a device such as a mobile entity (e.g., a mobile telecommunications device or a mobile phone or a user equipment or a terminal device).

[0026] According to a fourth exemplary aspect, a second method is disclosed. The method includes:

[0027] - Receiving (e.g., from the first device) a demodulation reference signal pattern on the resource grid, where the demodulation reference signal pattern is generated based on channel measurements of a radio channel between the first device and the second device measured on the resource grid.

[0028] The second method may include steps in which the device of the second exemplary aspect is configured to perform or has components for performing. The method may for example be performed and / or controlled by a device (e.g., a server, a server cloud or a RAN node or a part thereof).

[0029] Alternatively, the method according to any aspect may be performed and / or controlled by more than one device, such as a server cloud including at least two servers or a system of devices, such as a system including at least one RAN node and at least one UE. For example, the method may be performed and / or controlled by using at least one processor of the device. The method may include steps according to the third and fourth exemplary aspects and may for example be performed by a system of devices (e.g., a system including at least one RAN node and at least one UE).

[0030] According to a fifth exemplary aspect, a system is disclosed. The system includes a mobile entity or a part thereof and a server or a part thereof that together perform the method according to at least one of the third and fourth exemplary aspects.

[0031] The system may include one or more devices, such as the devices of any exemplary aspect (e.g., the first or second device), and / or a server, a server cloud, a RAN node, a UE, or a mobile terminal (e.g., a mobile telecommunications device or a mobile phone).

[0032] According to a sixth exemplary aspect, a computer program product is disclosed, which, when executed by a processor of a device, causes the device to perform a method according to at least one of the third exemplary aspect and the fourth exemplary aspect, or a method including at least one item:

[0033] - Obtain channel measurements of a radio channel between a first device and a second device measured on a resource grid;

[0034] - Generate (or derive) a demodulation reference signal pattern on the resource grid based on the channel measurements;

[0035] - Transmit (e.g., to the second device) the generated demodulation reference signal; and

[0036] - Receive (e.g., from the first device) a demodulation reference signal pattern on (the resource grid), wherein the demodulation reference signal pattern is generated based on channel measurements of a radio channel between the first device and the second device measured on the resource grid.

[0037] According to a further exemplary aspect, a device is disclosed, configured to perform and / or control, or include corresponding components for performing and / or controlling a method according to any exemplary aspect. According to a further exemplary aspect, a device is disclosed, including: at least one processor; and at least one memory storing instructions, which, when executed by the at least one processor, cause the first device to at least perform a method according to any exemplary aspect.

[0038] The disclosed device according to any aspect may be a module or component for a device (e.g., a chip). Alternatively, the disclosed device according to any aspect may be a device, such as a server, a server cloud, a RAN node, or a user equipment. The disclosed device according to any aspect may only include the disclosed components, such as parts, processors, memories, or may further include one or more additional components.

[0039] According to a further exemplary aspect, a computer program is disclosed, which, when executed by a processor, causes a device such as a server to perform and / or control actions of a method according to at least one of the third exemplary aspect and the fourth exemplary aspect.

[0040] According to a further exemplary aspect, a computer-readable storage medium is disclosed, which includes a computer program product according to the sixth exemplary aspect.

[0041] Any disclosure herein related to any exemplary aspect should be understood as being equally disclosed with respect to any subject matter according to the corresponding exemplary aspect, such as with respect to apparatus, methods, computer programs, and computer-readable media. Thus, for example, the disclosure of method steps should also be regarded as the disclosure of components for performing and / or causing the performance of the corresponding method steps. Similarly, the disclosure of components for performing and / or causing the performance of method steps should also be regarded as the disclosure of the method steps themselves. This also applies to any paragraph describing at least one processor; and at least one memory including instructions; the at least one memory and the instructions being configured to cause the apparatus to perform at least the steps in conjunction with the at least one processor.

[0042] Hereinafter, the exemplary features and exemplary embodiments of all aspects will be described in more detail.

[0043] According to all disclosed exemplary aspects, channel measurement can be understood as the measurement of channel characteristics, such as channel attenuation, noise power, fading, Doppler shift, Doppler spread, gain, K-factor, path loss. For example, channel measurements can be performed to obtain the channel frequency response. For example, channel measurements can be performed based on the demodulation reference signal pattern on the resource grid, and the successful decoding of the data transmitted on the resource grid along with the demodulation reference signal inserted after the demodulation reference signal pattern.

[0044] Channel measurements on a resource grid (e.g., a full-channel resource grid or a portion thereof) can be obtained, for example, using data-assisted channel estimation, or alternatively, by using reference (or pilot) signals to estimate the channel or a portion thereof, and by interpolating the channel between the reference (or pilot) signals. For example, the channel characteristics can be measured for resource elements in a resource grid with inserted DMRS, and interpolated or extrapolated to other resource elements of the resource grid. Through channel measurements on the resource grid, the channel frequency response can be obtained, and the demodulation reference signal pattern can be generated (or derived) on the resource grid based on this channel frequency response. The channel frequency response can represent the relationship between the transmitted signal and the received signal, such as the relationship between the input and output of a radio channel.

[0045] Channel recovery or estimation can be understood as reconstructing the channel frequency response on the resource grid. Channel recovery or estimation can be based on the signals received on resource elements that have been affected by channel characteristics (such as gain or noise). (Known) reference (or pilot) signals inserted into the resource elements of the resource grid can be used to recover or estimate the channel.

[0046] According to all disclosed exemplary aspects, a resource element may be understood as an element of a resource grid. For example, a resource element may represent a symbol in the time domain (e.g., OFDM) and a subcarrier in the frequency domain.

[0047] A resource grid may be a grid of resource elements. For example, a grid of 12 subcarriers and 14 OFDM symbols that results in 168 resource elements. The resource grid may be, for example, a full resource grid (of a radio channel) spanning the entire carrier bandwidth as defined in 3GPP TS 38.211, such as a time slot over the full frequency bandwidth of a radio channel, or the resource grid may be a subset of the full resource grid, for example, by applying some restrictions and / or decimation in the frequency domain and / or time domain to the full resource grid, such as by restricting channel measurements and DMRS pattern generation to a certain defined bandwidth and / or duration, or restricting to every, for example, two or four or eight PRBs and / or TTIs of the full resource grid, and / or restricting to one OFDM symbol out of every 2 or 4 OFDM symbols. For example, a RAN node may restrict a mobile entity to obtain or collect (channel measurements) only over a certain defined bandwidth and / or duration, for example, for training data collection. For example, a RAN node may perform decimation in the frequency domain and / or time domain, for example, in terms of bandwidth data collection complexity.

[0048] According to all disclosed exemplary aspects, a demodulation reference signal pattern may be a pattern of demodulation reference signals inserted into the resource grid (e.g., into the resource elements of the resource grid). A demodulation reference signal pattern may be a distribution of (multiple) resource elements in a resource grid associated with a radio channel (e.g., a physical resource block - transmission time interval (PRB - TTI) grid) that are designated for or carry (multiple) reference signals. The reference signal pattern may be dedicated to or customized for a radio channel, or may be specifically derived for a radio channel based on channel measurements obtained on the resource grid, and thus it may not be a predefined (or default or conventional) demodulation reference pattern, e.g., not a reference signal pattern predefined in a standard.

[0049] According to all disclosed exemplary aspects, the trained machine learning algorithm fed with channel measurements can be a machine learning algorithm that is fed with channel measurements on a resource grid. For example, the channel measurements can be used as an input to the machine learning algorithm. For example, a second device (e.g., gNB) can signal to a first device (e.g., UE) the default DMRS pattern for the training phase, e.g., by predefining DMRS patterns in the specification and configuring them or parts of them to the first device. During the ML algorithm training phase, the second device can transmit using the default DMRS pattern, and the first device can apply data-aided channel estimation to obtain channel measurements (or channel frequency responses) on the (e.g., full) channel resource grid or a part thereof, and use the channel measurements on the resource grid or a part thereof as training data to train the ML algorithm, or as test data to test the channel recovery performance of, e.g., the trained ML algorithm. The machine learning algorithm can be, for example, a convolutional neural network (CNN). Alternatively or additionally, deep learning techniques can be used to, for example, perform channel recovery or estimation. For example, by treating the resource grid as a 2-D image, the problem of channel recovery or estimation can be transformed into an image processing problem, e.g., similar to denoising or super-resolution. The training or test data for, e.g., the machine learning algorithm can be obtained using generated (e.g., standard-compliant) waveforms and generated channel models. This training data can be used to train the channel estimation CNN.

[0050] According to all disclosed exemplary aspects, the error performance metric can allow obtaining a measure of the error of the recovered channel relative to the original channel, e.g., the difference between the actual channel frequency response and the recovered or estimated channel frequency response. It can be or include a channel recovery performance metric that allows, for example, measuring the channel recovery performance, e.g., the mean square error (MSE) between the actual channel frequency response and the recovered or estimated channel frequency response. The error performance metric can be or include a bit error rate or a block error rate.

[0051] The channel recovery performance can be determined based on the error performance metric, e.g., using the channel frequency response.

[0052] According to all disclosed exemplary aspects, the nested demodulation reference signal pattern can be understood as a pattern arranged in a hierarchical structure. For example, if the patterns are in a sequence or serial construction, a pattern that is before another pattern in the sequence or has a lower index than another pattern can be included in the other pattern, e.g., w.r.t. the position of the resource element specifying or carrying the reference signal or data. The nested pattern can be obtained, for example, by an iterative method.

[0053] According to all disclosed exemplary aspects, a predetermined (or default or conventional) demodulation reference signal pattern may be a predetermined pattern from a standard, e.g., a DMRS pattern obtained using mapping type A or mapping type B, where mapping type A may start only at symbol 2 or symbol 3 within a time slot, and mapping type B always starts at the start of scheduling and the first symbol of the scheduling length indicator (SLIV). The DMRS transmission may be configured via higher layer signaling (RRC setup / reconfiguration), and once configured for a first device, the predetermined DMRS pattern may be fixed throughout the lifetime of the RRC connection.

[0054] According to all disclosed exemplary aspects, a change in a radio channel may be a change in at least one characteristic of the radio channel. For example, a change may be determined when the difference between a previous measurement of a characteristic and a new measurement of the characteristic exceeds a certain threshold, e.g., a difference of at least 50% of the characteristic may be determined as a change in the radio channel.

[0055] According to all disclosed exemplary aspects, a past demodulation reference signal pattern may refer to, for example, a demodulation reference signal generated or derived for a radio channel on an earlier occasion. It may be retrieved, for example, from the memory of a first device, and may be customized for the radio channel, for example, during a past communication session between the first device and a second device (e.g., by a method according to a third exemplary aspect).

[0056] According to an exemplary embodiment of the first exemplary aspect, a demodulation reference signal pattern is generated by using a trained machine learning algorithm fed with channel measurements.

[0057] According to an exemplary embodiment of the first exemplary aspect, the first device further comprises:

[0058] - means for receiving at least one item from a second device:

[0059] - a first information element that indicates a maximum number of resource elements in a resource grid to be used for insertion of the demodulation reference signal; and

[0060] - a second information element that indicates channel recovery performance associated with the demodulation reference signal pattern,

[0061] and wherein the demodulation reference signal pattern is further generated based on at least one of the first information element and the second information element.

[0062] According to an exemplary embodiment of the first exemplary aspect, the first device further comprises:

[0063] - means for generating a plurality of demodulation reference signal patterns on a resource grid based on channel measurements; and

[0064] - A component for sending multiple demodulation reference signal patterns to a second device,

[0065] wherein the multiple demodulation reference signal patterns correspond to different maximum numbers of resource elements in a resource grid where demodulation reference signals are to be inserted, or correspond to different channel recovery performances associated with corresponding demodulation reference signal patterns.

[0066] According to an exemplary embodiment of the first exemplary aspect, the first device further includes:

[0067] - A component for sending a third information element to the second device, the third information element indicating the channel recovery performance associated with the corresponding demodulation reference signal pattern among the multiple demodulation reference signal patterns.

[0068] According to an exemplary embodiment of the first exemplary aspect, the first device further includes:

[0069] - A component for sending multiple demodulation reference signal patterns as nested demodulation reference signal patterns.

[0070] This may, for example, allow for reducing signaling overhead.

[0071] According to an exemplary embodiment of the first exemplary aspect, channel measurements are performed based on a predetermined (or default or legacy) demodulation reference signal pattern and the successful decoding of data transmitted in a resource grid along with demodulation reference signals (e.g., data-aided channel estimation) inserted after the predetermined demodulation reference signal pattern.

[0072] According to an exemplary embodiment of the first exemplary aspect, the first device further includes:

[0073] - A component for sending a fourth information element to the second device, the fourth information element indicating the ability of the first device to generate demodulation reference signal patterns customized for a radio channel.

[0074] According to an exemplary embodiment of the first exemplary aspect, the first device further includes at least one of:

[0075] - A component for detecting a change in a radio channel, such as detecting a handover to a new radio channel;

[0076] - A component for triggering a fallback to a predetermined demodulation reference signal pattern (for the radio channel or the new radio channel);

[0077] - A component for generating a new demodulation reference signal pattern on a resource grid of the (radio channel) or a resource grid of the new radio channel based on new channel measurements performed after detecting a change in the radio channel, the new channel measurements being based on a predetermined demodulation reference signal pattern.

[0078] According to an exemplary embodiment of the first exemplary aspect, the first device further comprises:

[0079] - a component for sending one or more past demodulation reference signal patterns to a second device, the one or more past demodulation reference signal patterns being retrieved from a memory of the first device (e.g., a user equipment memory) and customized for a radio channel (e.g., during a past communication session between the first device and the second device).

[0080] One or more past demodulation reference signal patterns may be retrieved from the memory based on at least one of the following: an information element indicating the number of demodulation reference signal patterns to be generated or derived, an information element indicating the minimum number of resource elements to be used for the insertion of DMRS, an information element indicating the maximum number of resource elements to be used for the insertion of DMRS, an information element indicating the position of a (customized) DMRS pattern within a plurality of (e.g., nested) DMRS patterns, wherein the information element may be received from a device according to the second exemplary aspect.

[0081] According to an exemplary embodiment of the second exemplary aspect, the second device further comprises:

[0082] - a component for receiving a plurality of demodulation reference signal patterns from the first device,

[0083] wherein the plurality of demodulation reference signal patterns correspond to different maximum numbers of resource elements in a resource grid to be used for the insertion of demodulation reference signals, or correspond to different channel recovery performances associated with the respective demodulation reference signal patterns.

[0084] According to an exemplary embodiment of the second exemplary aspect, the second device further comprises:

[0085] - a component for receiving a third information element (or a plurality of third information elements) from the first device, the third information element indicating the (plurality of) channel recovery performances associated with the respective demodulation reference signal patterns among the plurality of demodulation reference signal patterns.

[0086] According to an exemplary embodiment of the second exemplary aspect, the second device further comprises:

[0087] - a component for selecting a demodulation reference signal pattern from the plurality of demodulation reference signal patterns to achieve a given error performance metric (e.g., bit error rate, block error rate, or channel recovery performance metric), wherein the error performance metric is determined based on at least one of the following:

[0088] - the effective number of resource elements for the insertion of the demodulation reference signal in the selected demodulation reference signal pattern;

[0089] - Channel recovery performance of the selected demodulation reference signal pattern; and

[0090] - Channel measurements of the radio channel.

[0091] According to an exemplary embodiment of the second exemplary aspect, the second device further comprises:

[0092] - Components for receiving multiple demodulation reference signal patterns as nested demodulation reference signal patterns (e.g., to reduce signaling overhead).

[0093] According to an exemplary embodiment of the second exemplary aspect, the second device further comprises:

[0094] - Components for sending to the first device at least one of the following:

[0095] - A first information element indicating the maximum number of resource elements in the resource grid to be used for the insertion of the demodulation reference signal; and

[0096] - A second information element indicating the channel recovery performance associated with the demodulation reference signal pattern,

[0097] And wherein the demodulation reference signal pattern is further generated based on at least one of the first information element and the second information element.

[0098] According to an exemplary embodiment of the second exemplary aspect, the second device further comprises:

[0099] - Components for receiving a fourth information element from the first device, the fourth information element indicating the ability of the first device to generate a demodulation reference signal pattern customized for the radio channel.

[0100] According to an exemplary embodiment of the second exemplary aspect, the second device further comprises:

[0101] - Components for receiving one or more past demodulation reference signal patterns from the first device, the one or more past demodulation reference signal patterns being retrieved from the memory of the first device (e.g., user equipment memory) and customized for the radio channel (e.g., during a past communication session between the first device and the second device).

[0102] According to an exemplary embodiment of the second exemplary aspect, the second device further comprises:

[0103] - Components for receiving a fourth information element from the first device, the fourth information element indicating the ability of the first device to generate a demodulation reference signal pattern customized for the radio channel.

[0104] According to an exemplary embodiment of the third exemplary aspect, the demodulation reference signal pattern is generated by using a trained machine learning algorithm that feeds channel measurements.

[0105] According to an exemplary embodiment of the third exemplary aspect, the first method further comprises:

[0106] - receiving (e.g., from a second device) at least one of the following:

[0107] - a first information element that indicates a maximum number of resource elements in a resource grid into which demodulation reference signals will be inserted; and

[0108] - a second information element that indicates a channel recovery performance associated with a demodulation reference signal pattern,

[0109] and wherein a demodulation reference signal pattern is further generated based on at least one of the first information element and the second information element.

[0110] According to an exemplary embodiment of the third exemplary aspect, the first method further comprises:

[0111] - generating a plurality of demodulation reference signal patterns on a resource grid based on channel measurements; and

[0112] - transmitting (e.g., to a second device) the plurality of demodulation reference signal patterns, wherein the plurality of demodulation reference signal patterns correspond to different maximum numbers of resource elements in a resource grid into which demodulation reference signals will be inserted, or correspond to different channel recovery performances associated with respective demodulation reference signal patterns.

[0113] According to an exemplary embodiment of the third exemplary aspect, the first method further comprises:

[0114] - transmitting (e.g., to a second device) a third information element that indicates a channel recovery performance associated with a respective demodulation reference signal pattern among the plurality of demodulation reference signal patterns.

[0115] According to an exemplary embodiment of the third exemplary aspect, the first method further comprises:

[0116] - transmitting (e.g., to a second device) the plurality of demodulation reference signal patterns as nested demodulation reference signal patterns (e.g., to reduce signaling overhead).

[0117] According to an exemplary embodiment of the third exemplary aspect, channel measurements are performed based on a predetermined (or default or legacy) demodulation reference signal pattern and successful decoding of data transmitted on a resource grid together with demodulation reference signals inserted after the predetermined demodulation reference signal pattern.

[0118] According to an exemplary embodiment of the third exemplary aspect, the first method further comprises:

[0119] - Transmit (e.g., to a second device) a fourth information element that indicates the ability of the first device to generate a demodulation reference signal pattern customized for a radio channel.

[0120] According to an exemplary embodiment of the third exemplary aspect, the first method further comprises at least one of the following:

[0121] - Detect a change in a radio channel, such as a handover to a new radio channel;

[0122] - Trigger a fallback to a predetermined demodulation reference signal pattern (for the radio channel or the new radio channel);

[0123] - Generate a new demodulation reference signal pattern on a resource grid of the (radio channel) or a resource grid of the new radio channel based on new channel measurements performed after detection of the change in the radio channel, the new channel measurements being based on a predetermined demodulation reference signal.

[0124] According to an exemplary embodiment of the third exemplary aspect, the first method further comprises:

[0125] - Transmit (e.g., to a second device) one or more past demodulation reference signal patterns retrieved from a memory of the first device (e.g., a user equipment memory) and customized for the radio channel (e.g., during a past communication session between the first device and the second device).

[0126] According to an exemplary embodiment of the fourth exemplary aspect, the second method further comprises:

[0127] - Receive (e.g., from a first device) a plurality of demodulation reference signal patterns, where the plurality of demodulation reference signal patterns correspond to different maximum numbers of resource elements in a resource grid where demodulation reference signals are to be inserted, or correspond to different channel recovery performances associated with the respective demodulation reference signal patterns.

[0128] According to an exemplary embodiment of the fourth exemplary aspect, the second method further comprises:

[0129] - Receive (e.g., from a first device) a third information element (or a plurality of third information elements) that indicates the (plurality of) channel recovery performances associated with the respective demodulation reference signal patterns among the plurality of demodulation reference signal patterns.

[0130] According to an exemplary embodiment of the fourth exemplary aspect, the second method further comprises:

[0131] - Select a demodulation reference signal pattern from multiple demodulation reference signal patterns to achieve a given error performance metric (e.g., bit error rate, block error rate, or channel recovery performance metric), where the error performance metric is determined based on at least one of the following:

[0132] - The effective number of resource elements for inserting the reference signal in the selected demodulation reference signal pattern;

[0133] - The channel recovery performance of the selected demodulation reference signal pattern; and

[0134] - Channel measurements of the radio channel.

[0135] According to an exemplary embodiment of the fourth exemplary aspect, the second method further includes:

[0136] - Receive (e.g., from a first device) multiple demodulation reference signal patterns as nested demodulation reference signal patterns (e.g., to reduce signaling overhead).

[0137] According to an exemplary embodiment of the fourth exemplary aspect, the second method further includes:

[0138] - Transmit (e.g., to a first device) at least one of the following:

[0139] - A first information element indicating the maximum number of resource elements in the resource grid that will be used for inserting the demodulation reference signal; and

[0140] - A second information element indicating the channel recovery performance associated with the demodulation reference signal pattern,

[0141] And wherein the demodulation reference signal pattern is further generated based on at least one of the first information element and the second information element.

[0142] According to an exemplary embodiment of the fourth exemplary aspect, wherein the second method further includes:

[0143] - Receive (e.g., from a first device) a fourth information element indicating the ability of the first device to generate a demodulation reference signal pattern customized for the radio channel.

[0144] According to an exemplary embodiment of the fourth exemplary aspect, the second method further includes:

[0145] - Receive (e.g., from a first device) one or more past demodulation reference signal patterns retrieved from the memory of the first device (e.g., user equipment memory) and customized for the radio channel (e.g., during a past communication session between the first device and the second device).

[0146] According to an exemplary embodiment of all exemplary aspects, the apparatus (e.g., the apparatus according to the first exemplary aspect) further comprises components for or the method further comprises: receiving, e.g., from a second apparatus, a configuration for a customized DMRS pattern, the configuration including at least one of the following: at least one default DMRS pattern (which may be used, e.g., for training a machine learning algorithm or for future transmissions), and a past demodulation reference signal pattern (which may be used, e.g., for training a machine learning algorithm or for future transmissions), wherein the past DMRS pattern may be retrieved, e.g., from a memory of the apparatus and may be customized for a radio channel during a past communication session between the apparatus and the second apparatus.

[0147] According to an exemplary embodiment of all exemplary aspects, the apparatus (e.g., the apparatus according to the first exemplary aspect) further comprises components for or the method further comprises: receiving, e.g., from a second apparatus, at least one default DMRS pattern; the apparatus generating, based on channel measurements, at least one (e.g., a plurality of) demodulation reference signal patterns on a resource grid, the channel measurements being performed, e.g., by the apparatus, e.g., by using a trained machine learning algorithm fed with the channel measurements. The channel measurements may be obtained or performed based on the received default (or predefined or conventional) demodulation reference signal pattern and the successful decoding of data transmitted on the resource grid with a demodulation reference signal (e.g., data-aided channel estimation) inserted after the predefined demodulation reference signal pattern. A customized (or specific) DMRS pattern may be generated using a machine learning algorithm, the DMRS pattern being customized to the radio channel.

[0148] According to an exemplary embodiment of all exemplary aspects, the apparatus (e.g., the apparatus according to the first exemplary aspect) further comprises components for or the method further comprises: sending, e.g., to a second apparatus, at least one (generated) demodulation reference signal pattern or a plurality of generated demodulation reference signal patterns, wherein the plurality of demodulation reference signal patterns corresponds to at least one of the following: different maximum numbers of resource elements in the resource grid where the demodulation reference signal is to be inserted, and different channel recovery performances associated with the respective demodulation reference signal patterns.

[0149] According to an exemplary embodiment of all exemplary aspects, the apparatus (e.g., the apparatus according to the first exemplary aspect) further comprises components for or the method further comprises: sending, e.g., to a second apparatus, a third information element together with a plurality of (generated, customized) DMRS patterns (e.g., in the same transmission), the third information element indicating the channel recovery performance associated with the respective demodulation reference signal pattern among the plurality of demodulation reference signal patterns. The plurality of demodulation reference signal patterns may be, e.g., a plurality of nested demodulation reference signal patterns.

[0150] According to an exemplary embodiment of all the exemplary aspects, the apparatus (e.g., the apparatus according to the first exemplary aspect) further comprises means for, or the method further comprises: receiving, e.g., from a second apparatus, a reconfiguration information element that indicates or comprises, e.g., at least one (customized) DMRS pattern from a plurality of demodulation reference signal patterns, the at least one (customized) DMRS pattern being, e.g., for future transmissions.

[0151] According to an exemplary embodiment of all the exemplary aspects, the apparatus (e.g., the apparatus according to the first exemplary aspect) further comprises means for, or the method further comprises: receiving, e.g., from a second apparatus, a control information element (e.g., DCI) that indicates or comprises a selected DMRS pattern from among the at least one of the plurality of demodulation reference signal patterns, the at least one selected DMRS pattern being, e.g., for future (scheduled) transmissions.

[0152] According to an exemplary embodiment of all the exemplary aspects, the apparatus (e.g., the apparatus according to the second exemplary aspect) further comprises means for, or the method further comprises: sending (dynamically), e.g., to a first apparatus, a control information element (e.g., DCI) that indicates or comprises the at least one selected DMRS pattern, the at least one selected DMRS pattern being, e.g., for future (scheduled) transmissions at least in part depending on a detected change in the radio channel (e.g., a handover to a different channel).

[0153] According to another exemplary embodiment, there is disclosed an apparatus (e.g., a UE or a gNB) that comprises means for obtaining a demodulation reference signal (DMRS) pattern specific to a radio channel using a machine learning algorithm, where the machine learning algorithm utilizes a resource grid of the radio channel (e.g., a full resource grid) and a channel frequency response of the radio channel measured on the resource grid or a portion thereof.

[0154] According to an exemplary embodiment of all the exemplary aspects, the apparatus further comprises components for or the method further comprises at least one of the following: applying a demodulation reference signal pattern according to a user equipment bit error rate requirement (e.g., for channel estimation or data demodulation); obtaining a user equipment bit error rate requirement; selecting a demodulation reference signal pattern according to a user equipment bit error rate requirement. According to an exemplary embodiment of all the exemplary aspects, the apparatus further comprises components for or the method further comprises at least one of the following: deriving (nested) multiple demodulation reference signal patterns based on an iterative method, sending an indication of the capability (e.g., to a gNB) to obtain a demodulation reference signal pattern specific to a radio channel; receiving (e.g., from a gNB) a parameter set for obtaining a demodulation reference signal pattern specific to a radio channel (e.g., the number of demodulation reference signal patterns to be obtained, the number of resource elements to be inserted for the demodulation reference signal); sending (e.g., to a gNB) a proposed demodulation reference signal pattern from a history based on the received parameter set and in particular a value indicating the channel recovery performance (e.g., mean square error) associated with the proposed demodulation reference signal pattern; obtaining multiple demodulation reference signal patterns based on the received parameter set, and optionally sending the obtained multiple demodulation reference signal patterns (e.g., to a gNB); receiving (e.g., from a UE) multiple demodulation reference signal patterns; selecting a demodulation reference signal pattern from the multiple demodulation reference signal patterns based on a bit error rate requirement; and sending (e.g., to a UE) a demodulation reference signal pattern from the multiple demodulation reference signal patterns to be applied to communication on a radio channel. According to an exemplary embodiment of all the exemplary aspects, the apparatus further comprises components for the method or the method further comprises at least one of the following: determining a handover from a radio channel to a second radio channel; when it is determined that there is a handover from a radio channel to a second radio channel, obtaining a second demodulation reference signal pattern specific to the second radio channel using a machine learning algorithm.

[0155] The above features and example embodiments may be described with respect to DMRS patterns. However, the techniques described for (generating) DMRS patterns may also be used for (e.g., deriving) CSI-RS or SRS patterns using similar signaling and techniques.

[0156] The above features and example embodiments may equally relate to different aspects, and the demodulation reference signal pattern is associated with at least one of the following: the number indicating the number of resource elements (per time slot) used for the demodulation reference signal in the demodulation reference signal pattern; and a value indicating the channel recovery performance (e.g., mean square error) associated with the demodulation reference signal pattern.

[0157] It should be understood that the presentation in this section is exemplary only and not restrictive.

[0158] Other features will become apparent from the following detailed description considered in conjunction with the accompanying drawings. It should be understood, however, that the drawings are designed solely for the purpose of illustration and not as a definition of the limits, for which reference should be made to the appended claims. It should also be understood that the drawings are not drawn to scale and that they are merely intended to conceptually illustrate the structures and processes described herein. BRIEF DESCRIPTION OF THE DRAWINGS

[0159] Figure 1 is a schematic diagram showing an example system;

[0160] Figure 2 is a flowchart showing an example embodiment of a method according to a third exemplary aspect;

[0161] Figure 3 is a flowchart showing an example embodiment of a method according to a fourth exemplary aspect;

[0162] Figure 4a shows an example of a DMRS pattern;

[0163] Figure 4b shows an example of a DMRS pattern;

[0164] Figure 5 is a flowchart showing an exemplary embodiment of a method according to at least one of the third and fourth exemplary aspects;

[0165] Figure 6 is an exemplary embodiment of a method according to at least one of the third and fourth exemplary aspects;

[0166] Figure 7 is a flowchart showing an exemplary embodiment of a method according to at least one of the third and fourth exemplary aspects;

[0167] Figure 8a shows an example of a DMRS pattern according to any exemplary aspect.

[0168] Figure 8b is a table having characteristics associated with Figure 8a the exemplary DMRS pattern, particularly the number of resource elements in the resource grid for the insertion of the demodulation reference signal and the channel recovery performance.

[0169] Figure 9a shows an exemplary DMRS pattern with DMRS inserted into 12 resource elements;

[0170] Figure 9b shows a diagram of an exemplary DMRS pattern with DMRS inserted into 12 resource elements according to any exemplary aspect;

[0171] Figure 10 Schematic block diagram of an exemplary embodiment of an apparatus according to a first aspect and / or an exemplary embodiment of an apparatus configured to perform a method according to a second exemplary aspect. Detailed implementation

[0172] The following description is for better understanding and should be understood as supplementing and read together with the description provided in the above overview section of this specification. Some aspects may have different terms from those provided in the above description, for example. However, those skilled in the art will understand that these terms refer to the same subject matter more specifically. For example, generating a DMRS pattern may be referred to as deriving a DMRS pattern, a customized or specific DMRS pattern may be referred to as a flexible DMRS pattern, the channel frequency response on a resource grid may be referred to as a channel grid, and multiple DMRS patterns may be referred to as a set of DMRS patterns.

[0173] Figure 1 Schematic diagram showing an exemplary system. Among them, a RAN node (e.g., gNB) 101 or a base station 101 sends to a mobile entity 102 (e.g., a user equipment 102) via a DL channel 103 and receives via a UL channel 104. For example, DL transmission may utilize a default DMRS pattern or a customized DMRS pattern, such as for demodulation or decoding of a data channel. The exemplary system may be, for example, a system according to a sixth exemplary aspect. The exemplary system may, for example, execute a method of at least one of the third and fourth exemplary aspects. The exemplary system may include an exemplary embodiment of a device according to a first aspect and an exemplary embodiment of a device according to a second aspect.

[0174] Figure 2 Flowchart showing an exemplary embodiment of a method according to a third exemplary aspect; the flowchart 200 may be executed, for example, by at least one of the devices according to a first exemplary aspect (e.g., a mobile entity or a UE 102).

[0175] In a first step 201, obtain or measure channel measurements of a radio channel between a first device (e.g., a UE 102) and a second device (e.g., a gNB 101 or a base station 101) on a resource grid, for example, on the entire or all of the resource grid or a subset of the entire resource grid.

[0176] In a second step 202, generate or derive a demodulation reference signal pattern on the resource grid based on the channel measurements.

[0177] In a third step 203, send the generated demodulation reference signal pattern to the second device.

[0178] Figure 3is a flowchart showing an exemplary embodiment of a method according to a fourth exemplary aspect; the flowchart 303 may be executed, for example, by at least one of the devices according to the second exemplary aspect (such as the base station 101 or the gNB 101).

[0179] In step 301, a demodulation reference signal pattern on a resource grid is received, for example, from a first device (such as the UE 102), where the demodulation reference signal pattern is generated based on channel measurements of a radio channel between the first device and a second device measured on the resource grid.

[0180] Figure 4a and Figure 4b shows exemplary predetermined or conventional DMRS patterns for different antenna ports. In each case, a resource grid with 14 OFDM symbols and 12 subcarriers is shown, where the resource element 402 carries the DMRS or is used for inserting the DMRS, the resource element 403 carries the payload (such as data), and the resource element 401 is empty. In Figure 4a the DRMS has been inserted at OFDM symbol 2 and OFDM symbol 3. In Figure 4b additionally, the DRMS has been inserted at OFDM symbol 10 and OFDM symbol 11.

[0181] Figure 5 is a flowchart showing an exemplary embodiment of a method according to at least one of the third and fourth exemplary aspects; the exemplary method may be executed by a system, in particular by the gNB 101 and the UE 102 together.

[0182] During the RRC establishment 501, the UE 102 may send a fifth information element 502 to the gNB 101, which indicates the ability of the first device (e.g., the UE 102) to generate demodulation reference signal patterns customized for a radio channel. The fifth information element may also indicate, for example, the ability of the first device (e.g., the UE 102) to send one or more past demodulation reference signal patterns, and specifically, to send one or more past demodulation reference signal patterns retrieved from the memory of the UE 102 (e.g., the user equipment memory) and customized for the radio channel during a past communication session between the UE 102 and the gNB 102. The one or more past demodulation reference signal patterns may be retrieved from the memory based on at least one of the following: an information element indicating the number of DMRS patterns to be derived, an information element indicating the minimum number of resource elements to be used for the insertion of the DMRS, an information element indicating the maximum number of resource elements to be used for the insertion of the DMRS, an information element indicating the position of the customized DMRS pattern within a plurality of DMRS patterns, where the information elements may be received from the gNB 102. For example, when the UE 102 or another (e.g., similar) mobile entity was located in the same (e.g., geographical) location in the past, e.g., in a fixed wireless access use case, one or more past demodulation reference signal patterns may have been generated. The UE 102 may send, for example, an information element indicating the channel recovery performance associated with the respective demodulation reference signal pattern to the gNB 101 together with the one or more past demodulation reference signal patterns.

[0183] UE 102 may receive 503 a configuration for a customized DMRS pattern from gNB 101, where the customized DMRS pattern includes at least one of the following: at least one default DMRS pattern that may be used for training a machine learning algorithm or for future transmissions 504 (e.g., as a fallback option), and past demodulation reference signal patterns that may be used for training a machine learning algorithm or for future transmissions 504, where the past DMRS patterns may be retrieved from the memory of the UE or gNB and are customized, for example, for a radio channel during a past communication session between UE 102 and gNB 101 (e.g., the DMRS pattern transmitted 502 by UE 102 to gNB 101). Receiving 503 at least one default DMRS pattern may enable UE 102 to generate at least one (e.g., multiple) demodulation reference signal patterns on a resource grid based on channel measurements, e.g., performed by UE 102, e.g., via a trained machine learning algorithm fed with channel measurements. Channel measurements may be obtained or performed based on the received default (or predefined or legacy) demodulation reference signal pattern and the successful decoding of data transmitted on the resource grid together with a demodulation reference signal (e.g., data-aided channel estimation 505) inserted after the predefined demodulation reference signal pattern. A machine learning algorithm may be utilized 506 to generate the customized DMRS pattern (see, e.g., Figure 6 and Figure 7 ). gNB 101 may signal to UE 102 the default DMRS pattern for a training phase, e.g., by predefining the DMRS pattern in a specification and configuring some for the UE. During the ML model training phase, the gNB may transmit using the default DMRS pattern, the UE may apply data-aided channel estimation to obtain the channel frequency response on the resource grid, and may use the channel frequency response on the resource grid as training data to train the ML model. Alternatively or additionally, when the UE channel changes due to relocation or beam update, the DMRS pattern may be updated accordingly, i.e., the gNB may fallback to the default DMRS pattern and trigger the UE to retrain the model and feedback a new set of generated or customized DMRS patterns. Further, the gNB may configure the UE to use only some or all of these patterns signaled by parameters derived from the DMRS pattern determined or configured at the gNB by the UE. Alternatively or additionally, the gNB may configure the UE to use the default pattern for future transmissions 504, which may implicitly trigger the training of the ML model.

[0184] During RRC reconfiguration 507, the UE 102 may send 508 to the gNB 101 at least one generated demodulation reference signal pattern or a plurality of demodulation reference signal patterns, where the plurality of demodulation reference signal patterns corresponds to at least one of the following: different maximum numbers of resource elements in the resource grid where the demodulation reference signal is to be inserted, and different channel recovery performances associated with the respective demodulation reference signal patterns. The UE 102 may, for example, send to the gNB 101 together with a plurality of (generated, customized) DMRS patterns a third information element indicating the channel recovery performance associated with the respective demodulation reference signal pattern among the plurality of demodulation reference signal patterns. The plurality of demodulation reference signal patterns may, for example, be a plurality of nested demodulation reference signal patterns. The UE 102 may receive 509 from the gNB 101 reconfiguration information that indicates or includes, for example, at least one (customized) DMRS pattern from among the plurality of demodulation reference signal patterns for future transmissions.

[0185] gNB 101 may select at least a demodulation reference signal pattern from a plurality of demodulation reference signal patterns to achieve a given error performance metric (e.g., bit error rate) and, for example, minimize the RS overhead, where the error performance metric may be determined based on, for example, at least one of the following: the number of resource elements with valid insertions of demodulation reference signals in the selected demodulation reference signal pattern; the channel recovery performance of the selected demodulation reference signal pattern; and the channel measurements of the radio channel. UE 102 may receive control information (e.g., DCI) from gNB 101 that indicates or includes, for example, at least one selected DMRS pattern to be used for future (scheduled) transmissions from the plurality of demodulation reference signal patterns. gNB 101 may (dynamically) send control information (e.g., DCI) to UE 102 that indicates or includes at least one selected DMRS pattern, for example, at least in part depending on detecting a change in the radio channel (e.g., switching to a different channel), for example, for future (scheduled) transmissions. The gNB may instruct the UE to switch between the generated DMRS patterns (e.g., from the generated plurality of demodulation reference signal patterns). The gNB may dynamically schedule the (generated) DMRS patterns and may indicate the selected DMRS pattern for the scheduled transmission in the DCI. As an exemplary embodiment, this may be done by creating a new field for (flexible) DMRS pattern indication in the DCI format. The bit width of the field may be consistent with the number of generated (flexible) DMRS patterns, and when a bit in the field is set to "1", it may indicate that the associated DMRS pattern is selected for transmission. Alternatively, the gNB may encode the (logical) identifier of the selected (flexible) DMRS pattern in the DCI or use an alternate identifier not used by the default DMRS pattern. Additionally or alternatively, the UE may signal to the gNB in the UCI the preferred DMRS pattern (e.g., from the generated plurality of demodulation reference signal patterns) for the next DL transmission (e.g., UE-assisted (flexible) DMRS pattern transmission). This may be done by creating a new field in the UCI format for flexible DMRS pattern reporting. The gNB may then select the DMRS pattern for a later DL transmission, for example, considering the preferred pattern signaled by the UE.

[0186] The gNB may instruct the UE to fallback to the default DMRS pattern. For example, the gNB may collect UE ACK / NACK feedback and may monitor the UE's BER statistics. When the BER statistics deviate from the expected value, the gNB may use DCI to send a signaling notification to the UE to switch to the default DMRS pattern, which may implicitly trigger the retraining of the ML model. As an exemplary embodiment, if all flexible DMRS patterns are disabled, where the corresponding index signaling notification is "0", it may implicitly indicate a fallback to the default DMRS pattern. Note that the proposed method may allow using the number of DMRS REs and the associated channel recovery MSE to label the DMRS pattern, or allow using any loss function (and correspondingly the threshold associated with the pattern) to operate and label the DMRS pattern. Another note is that the DMRS pattern may be configured non-uniformly across the bandwidth to achieve a specific BER / MSE target between the BER / MSE values in the DMRS pattern set. Thus, for example, more bits may be used to indicate the DMRS pattern of the DCI format (e.g., PRB-specific DMRS pattern scheduling)

[0187] The DMRS pattern can be selected to meet the target channel recovery MSE requirement and can be scheduled for transmission to minimize the RS overhead. The derived DMRS pattern can be directly scheduled to meet the bit error rate (BER) requirement of the UE. The BER can be calculated using the DMRS RE number, channel recovery MSE, UE measurements (e.g., UE channel attenuation, noise power, etc.). For each of the multiple (generated) DMRS patterns, the BER can be calculated, and then one pattern with the minimum RS overhead can be selected to meet the BER requirement and can be used for transmission.

[0188] The BER can be the BER performance after the equalizer, which can be a function of the SNR after the equalizer, and can be calculated as a function of the MSE of the equalizer output, as shown below:

[0189]

[0190] where Q(x) can be defined as and can depend on the modulation scheme. J may represent the MSE of the equalizer output. The MSE of the equalizer output can consist of two parts, the MSE of the equalizer assuming perfect channel knowledge and the MSE of the channel recovery MSE,

[0191]

[0192] where, J ch can represent the channel recovery MSE, σ 2 represents the noise power, H l, where \(l = 1,\cdots,M\) represents the channel attenuation at the \(l\)-th allocated PRB. The channel attenuation can be derived using UE CSI reports, such as per-PRB CQI, and the noise power can be derived using UE measurement reports, such as RSSI and RSRQ. Therefore, the gNB can calculate the BER performance of each DMRS pattern in the derived DMRS pattern before scheduling the DMRS pattern for transmission. To meet the BER performance requirements of a specific UE, the gNB can select the DMRS pattern with the maximum available MSE below the BER requirement limit, thereby minimizing, for example, the DMRS overhead for transmission. In another way, since the UE can know all the measurements it has performed, the UE can directly calculate the BER performance of each DMRS pattern in the configured DMRS pattern and can indicate its preferred DMRS pattern in the UL UCI to assist the gNB in flexible DMRS pattern scheduling.

[0193] For example, the UE can first report its ability to support the generation of demodulation reference signal patterns customized for the radio channel. The gNB can control the generation or derivation of the DMRS pattern by the UE (e.g., according to the number of DMRS patterns to be determined and the corresponding maximum number of DMRS REs for thresholding). Next, the UE can report multiple DMRS patterns customized for the actual UE radio channel with the minimum DMRS overhead. Each reported DMRS pattern can be associated with a different MSE threshold. Traditional or default DMRS patterns can be used during the training phase. The UE can apply data-aided channel estimation to estimate, recover, or derive (e.g., complete) the channel frequency response on the resource grid and can use the derived channel frequency response on the resource grid as training data to train the ML model. Then, the gNB can dynamically select one of the reported DMRS patterns for DL communication with the UE to meet a given BER target. The BER can be estimated based on the reported MSE, the number of DMRS REs in the DMRS pattern, and UE measurements. The (optimal) DMRS pattern for UL communication can also be derived or learned at the gNB side, and if UL / DL channel reciprocity is applicable (e.g., TDD communication), the (optimal) DMRS pattern for DL communication can also be derived or learned. In this case, for example, some of the proposed signaling may still apply (e.g., new UE capability reports, sending multiple (optimal) DMRS patterns to the UE).

[0194] Once configured during RRC connection establishment, the DMRS pattern can be fixed for the UE, for example. That is, even if the UE channel may change during that time, the configured DMRS pattern can be used throughout the life cycle of the RRC connection. Alternatively (e.g., instead of fixing a set of DMRS patterns throughout the life cycle of the RRC connection, or instead of a predefined set of DMRS patterns and applying them to all UEs, for example), the DMRS pattern can be generated or derived as described and applied dynamically, for example, when a change in the radio channel is detected, to meet the UE's BER requirements and / or minimize the DMRS overhead. The DMRS pattern can be adaptively changed (flexibly) based on, for example, the UE's BER / MSE performance requirements. Since the DMRS pattern can achieve or is achieving the maximum available MSE to meet the BER performance requirements, the DMRS overhead for DL transmission can be minimized.

[0195] For a set of DMRS RE number thresholds (e.g., indicated by a first information element indicating the maximum number of resource elements in the resource grid where demodulation reference signals will be inserted for demodulation), the proposed ML model can output a nested set of DMRS patterns according to the DMRS RE number threshold, where each of the nested DMRS patterns can have different channel recovery MSE performance. Note that the same ML model can be used to recover (estimate or predict) the channel. For each DMRS pattern in the derived DMRS patterns, the BER of the UE channel can be estimated using the DMRS RE number, the communication party channel recovery MSE, and the UE measurement. Therefore, the generated (optimal) DMRS pattern with the minimum RS overhead can be selected from the pattern set to meet the target BER requirement and used for transmission.

[0196] The DMRS transmission framework can be updated to accommodate the proposed DMRS pattern derivation and (flexible) DMRS scheduling methods. Since the DMRS pattern generated or derived on the resource grid based on channel measurements (e.g., from a real UE channel) can reflect the true correlation between the resource elements within the channel frequency response (in both time domain and frequency domain), the pattern can appear random. Therefore, the derived pattern can be nested, and signaling the nested random pattern can minimize the signaling overhead.

[0197] For the proposed ML model-based method, use cases can include, for example, both DL and UL DMRS pattern derivation and (flexible) DMRS scheduling. The proposed method can be implemented at the RAN node (e.g., gNB) or the UE side.

[0198] Figure 6FIG. is a diagram illustrating an exemplary embodiment of a method according to at least one of the third and fourth exemplary aspects, specifically illustrating the generation of DMRS patterns by means of a trained machine learning algorithm 602 and the generation of a plurality of nested demodulation reference signal patterns 601, 602, 607. The REs 604 into which DMRS is inserted are indicated by crosses, e.g., RS RE 604. The REs 605 without DMRS inserted (e.g., data REs 605) are indicated by diagonals. The channel frequency response on the resource grid (e.g., channel grid) has been obtained, for example, by data-aided channel estimation. Since all resource elements may have been measured or estimated, they can all be considered to carry or have inserted DMRS (because they can all be recovered, for example, by a machine learning algorithm). In a first step, a first DMRS pattern 601 can be generated by selecting an RE and removing the DMRS therefrom. The ML algorithm 602 can then recover or estimate the RE selected in the channel frequency response 603 on the resource grid. This can be done continuously for all REs. For each DMRS pattern in the generated DMRS patterns having one RE without DMRS, the channel recovery performance associated with the corresponding demodulation reference signal pattern can be determined, for example, by determining the MSE (value) between the training channel frequency response (e.g., training channel grid) on the resource grid and the corresponding recovered channel frequency response (e.g., recovered channel grid) on the resource grid, which can be recovered based on the generated DMRS pattern, for example, using a (trained) ML algorithm. The DMRS pattern 601 having an RE 605 with DMRS that does not exhibit the best channel recovery performance can be selected as the basis for continuously removing DMRS from a second RE in the next step to generate a DMRS pattern 606 having two REs without DMRS (e.g., two data REs 605). The process can then be iteratively repeated (e.g., via a loop), for example, until all REs are data REs or a predetermined number of REs in the resource grid are used for data or DMRS, or until a predetermined channel recovery performance (threshold) can no longer be achieved (e.g., because too many REs are data REs). An example of such a process is shown in Figure 7 which is a flowchart illustrating an example embodiment of a method according to at least one of the third and fourth exemplary aspects, specifically illustrating the generation of DMRS patterns by a trained machine learning algorithm and the generation of a plurality of nested demodulation reference signal patterns. The plurality of (generated) DMRS patterns can be a series of nested patterns and can be represented, for example, in a nested manner, which can allow, for example, minimizing the signaling overhead.

[0199] Multiple (nested) DMRS patterns can be generated by a method with a recurrent ML model, e.g., based on channel measurements (e.g., from UE channels), e.g., at different DMRS RE number thresholds. For example, according to different DMRS RE number thresholds, the derived DMRS patterns can have different channel recovery MSE performances.

[0200] Since the channel frequency response on the resource grid (e.g., channel PRB-TTI grid) can be regarded as an image with, e.g., 14 (number of symbols in the time domain) × 12 (number of subcarriers in the frequency domain) pixels, machine learning algorithms such as convolutional neural networks (CNNs) can be used for image processing. For example, an extended convolutional neural network model can be used to derive DMRS patterns from the channel (measurements) for different overhead parameters (e.g., different numbers of DMRS RE inserted into the resource grid).

[0201] Training the proposed ML model can utilize, e.g., the channel frequency response on a (full) resource grid with, e.g., 14 × 12 pixels (e.g., full channel grid) as training data. Data-aided channel estimation can be used to collect this data, which can be used to improve channel estimation performance. When the transport block is successfully decoded, the decoded data REs can be regarded as pilots. Using the decoded data REs together with the reference signal REs, the channel frequency response of the entire transport block can be obtained, and thus the full channel frequency response on the resource grid can be obtained.

[0202] As Figure 6 shown, it can start from a pattern where all REs are DMRS REs, so the full channel frequency response on the resource grid can be available and the channel can be recovered. Then, for example, one RE that is suppressed (e.g., setting the channel at the corresponding RE to 0) is picked from the DMRS transmission, and, e.g., the CNN is trained to use the remaining DMRS REs to recover the channel, and the remaining DMRS REs have a loss function defined as the MSE between the recovered channels.

[0203] Training a machine learning algorithm to recover a radio channel can be based on a loss function defined as the MSE between the recovered channel frequency response (e.g., the recovered channel) and the original (or true) channel frequency response H = {h i，j}(e.g., the original channel).

[0204]

[0205] Given the channel frequency response H, h i，j = 0 can be assigned to mimic the suppression of DMRS transmission on the RE at the i-th row and j-th column, and H' is obtained. H' (with hi,j The channel data set with H) = 0 is fed into a machine learning algorithm, and the MSE can be used as the loss function to train the machine learning algorithm to recover the channel frequency response on the resource grid. When the training is completed, the machine learning algorithm I can be tested with a test data set H″ (with another batch of channel samples H with h i,j = 0), and the MSE of RE(i, j) can be recorded. This step can be repeated for all REs. Then, the RE with the minimum MSE can be selected, which can be marked as the data RE, and the RE can be recorded as, for example, a (index, MSE) pair, as a DMRS pattern with 12×14 - 1 DMRS REs, with the corresponding channel recovery MSE value. The entire process can be repeated for the remaining DMRS REs (the selected RE is marked as the data RE, where the marking of the data RE is incremental, for example). The model can be retrained, the MSE values can be recorded, and one RE with the minimum MSE can be found, marked, and recorded, for example, until all REs are suppressed. A series of 12×14 DMRS patterns with increasing channel recovery MSE can be obtained. The trained models for each of the selected DMRS patterns can be recorded and can be used, for example, to predict / estimate the channel when scheduling the corresponding DMRS pattern.

[0206] Note that since the channel frequency response "image" can have (only) 12×14 pixels, the CNN utilized in the proposed method can be very simple, and thus the training process can be lightweight. The trained models for each of the selected DMRS patterns can be recorded and the trained models can be used, for example, to predict / estimate the channel when using the corresponding DMRS pattern.

[0207] In addition, the machine learning algorithm can include or utilize deep learning techniques to, for example, perform channel estimation or recovery. For example, by treating the resource grid as a 2-D image, the problem of channel estimation can be transformed into an image processing problem, similar to denoising or super-resolution, where a CNN can be effective. Standard-compliant waveforms and channel models can be customized and generated for the training data. Using deep learning, this training data can be used to train the channel estimation CNN. This example can show how to generate such training data and how to train the channel estimation CNN. This example can also show how to use the channel estimation CNN to process an image containing the channel frequency response obtained using (for example, linearly interpolated) received pilot symbols.

[0208] Figure 8a and Figure 8bShows examples of (generated) DMRS patterns 801, 802, 803, 804, 805 according to any exemplary aspect. The exemplary DMRS patterns 801, 802, 803, 804, 805 may be nested because, for example, the REs without DMRS in pattern 801 are also the REs without DMRS in (nested) patterns 802, 803, 804, 805; the REs without DMRS in pattern 802 are also the REs without DMRS in (nested) patterns 803, 804, 805; the REs without DMRS in pattern 803 are also the REs without DMRS in (nested) patterns 804, 805; and the REs without DMRS in pattern 804 are also the REs without DMRS in (nested) pattern 805. Alternatively, to nest the patterns based on the REs without DMRS or data REs, the patterns may be nested relative to the REs with DMRS (e.g., by reversing Figure 8a the order of the DMRS patterns in) or any other nesting criterion. Figure 8b The rows in the table in Figure 8b correspond to the exemplary DMRS patterns 801, 802, 803, 804, 805. For example, for the DMRS pattern 801 with 159 REs with DMRS, an MSE value of 0.06 can be determined, and for the DMRS pattern 805 with 35 REs with DMRS, a (higher) MSE value of 0.1 can be determined.

[0209] The proposed method can be tested using an aligned channel model with a mixture of scenarios and UE speeds to evaluate the performance gain in a general case, specifically, for example, the channel dataset parameters shown below:

[0210] Channel: 612 subcarriers x 256 TTI for SISO channel

[0211] delayProfiles = {'TDL-A', 'TDL-B', 'TDL-C', 'TDL-D', 'TDL-E'};

[0212] channel.DelayProfile = string(delayProfiles(randi([1 numel(delayProfiles)])));

[0213] channel.DelaySpread = randi([1 300]) * 1e-9;

[0214] channel.MaximumDopplerShift = randi([5 400]);

[0215] The derived patterns are asFigure 8a As shown, the striped pixels represent the data REs marked as RE, and the cross pixels represent DMRS REs. It can be seen that the derived DMRS pattern is very different from, for example, the DMRS pattern predefined in the 3GPP standard, and the positions of the DMRS REs can seem random. As more REs are marked as data REs and the REs reserved for DMRS transmission decrease, the channel recovery MSE can increase (e.g., Figure 8b ).

[0216] Figure 9a is a diagram showing an exemplary predefined (or default) DMRS pattern with DMRS inserted in 12 resource elements, which can be compared with an exemplary (generated) DMRS pattern showing DMRS inserted in 12 resource elements 902 according to any exemplary aspect. Figure 9b Both DMRS patterns have the same number of REs with DMRS, i.e., 12 DMRS REs 602 (the rest are, for example, data REs 901). However, in this example, the MSE of the generated DMRS pattern is 0.0068, while the MSE for the predefined pattern is 0.018. Therefore, the generated DMRS pattern can be considered to perform better than the predefined DMRS pattern. Note that these test results are for the general case, and for the UE-specific case, the performance gain may be even greater.

[0217] Since the generated or derived DMRS patterns can be, for example, a series of nested patterns, they can be represented in a nested manner to, for example, minimize the signaling overhead. For example, if the UE sends 5 patterns from the derived patterns based on the gNB-configured parameters, and they have 35, 50, 67, 109, 159 DMRS REs respectively. The first DMRS pattern with 35 DMRS REs can be represented and signaled with 12×14 bits, where "1" represents a DMRS RE and "0" represents a data RE (or vice versa). The second DMRS pattern may only require 12×14 - 35 bits to be represented and signaled because the derived patterns can be nested, and only those data REs in the first pattern will be further marked as DMRS for the second pattern. Therefore, this nested characteristic of the derived DMRS patterns can be used to reduce the pattern signaling overhead. Note that this may depend on the selected flexible DMRS pattern, and there may be other ways to further compress the signaling overhead.

[0218] Figure 10 is a schematic block diagram of an apparatus 1000 according to an exemplary aspect, which can, for example, represent one of the apparatuses according to the first or second exemplary aspect.

[0219] The apparatus 600 includes a processor 1001, a program memory 1002, a working or main memory 1003, a data memory, one or more communication interfaces 1004, and an optional user interface 1005.

[0220] The apparatus 1000 can be configured, for example, to perform and / or control or include corresponding components (at least one of components 1001 to 1005) for performing and / or controlling the methods according to the third and / or fourth exemplary aspects. The apparatus 1000 can also constitute an apparatus including at least one processor (1001) and at least one memory (1002) storing instructions, which when executed by the at least one processor cause the apparatus (e.g., the apparatus 1000) to at least perform and / or control the methods according to all exemplary aspects.

[0221] The processor 1001 can further control, for example, the memories 1002 to 1003, the one or more communication interfaces 1004, and the optional user interface 1005.

[0222] The processor 1001 can execute, for example, program code stored in the program memory 1002, and the program memory 1002 can represent a readable storage medium including the program code, which when executed by the processor 1001 causes the processor 1001 to execute the methods according to the third or fourth exemplary aspects.

[0223] The processor 1001 (and any other processor mentioned in this specification) can be any suitable type of processor. The processor 1001 can include, but is not limited to, one or more microprocessors, one or more processors with one or more accompanying digital signal processors, one or more processors without accompanying one or more digital signal processors, one or more dedicated computer chips, one or more field programmable gate arrays (FPGAs), one or more controllers, one or more application specific integrated circuits (ASICs), or one or more computers / servers. The relevant structure / hardware has been programmed in such a way as to implement the described functions. The processor 1001 can be, for example, an application processor running an operating system.

[0224] The program memory 1002 may also be included in the processor 1001. The memory may be, for example, fixedly connected to the processor 1001 or at least partially removable from the processor 1001, for example, in the form of a memory card or a memory stick. The program memory 1002 may be, for example, a non-volatile memory. For example, it may be a flash memory (or a part thereof), any ROM, PROM, EPROM, and EEPROM memory (or a part thereof), or a hard disk (or a part thereof), to name but a few examples. The program memory 1002 may also include an operating system for the processor 1001. The program memory 1002 may also include firmware for the device 1000.

[0225] The device 1000 includes a working memory 1003, for example, in the form of a volatile memory. For example, it may be a random access memory (RAM) or a dynamic RAM (DRAM), to give but a few non-limiting examples. It may be used, for example, by the processor 1001 when executing the operating system and / or a computer program.

[0226] The data memory may be, for example, a non-volatile memory. For example, it may be a flash memory (or a part thereof), any ROM, PROM, EPROM, and EEPROM memory (or a part thereof), or a hard disk (or a part thereof), to name but a few examples. The data memory may store, for example, one or more pieces of information, such as a measurement of a first signal, information about a first subspace, additional information.

[0227] (A plurality of) communication interfaces 1004 enable the device 1000 to communicate with other entities (for example, another device). (A plurality of) communication interfaces 1004 may include, for example, a wireless interface (such as a cellular radio communication interface and / or a WLAN interface) and / or a wired interface (such as an IP-based interface) to communicate with an entity via the Internet, for example. (A plurality of) communication interfaces may enable the device 1000 to communicate with other entities (for example, one or more entities included in a mobile communication network).

[0228] The user interface 1005 is optional and may include a display for displaying information to the user and / or an input device (such as a keyboard, keypad, touchpad, mouse, etc.) for receiving information from the user.

[0229] Some or all of the components of the device 1000 may be connected via a bus, for example. Some or all of the components of the device 1000 may be combined into one or more modules, for example.

[0230] The disclosed exemplary aspects may allow for enhanced signaling, efficient resource reservation, and interoperability. The generated or derived DMRS patterns may allow for reduced overhead, e.g., for signaling patterns and / or to achieve a trade-off between DMRS overhead and channel recovery MSE. The exemplary aspects and exemplary embodiments provided below may implement adaptive scheduling of DMRS patterns, e.g., based on the UE's BER / MSE performance requirements.

[0231] A method (or apparatus) according to any aspect, particularly for generating a demodulation reference signal pattern on a resource grid based on channel measurements, may be used for or extended to multiple layers, multiple users, and multiple antenna panels, e.g., by converting a 2-D CNN network into a 3-D CNN network (the additional dimension representing multiple layers, multiple users, or multiple antenna panels). Instead of using a 2D image of the channel frequency response on the resource grid, a machine learning algorithm may use or learn from a cube of channel matrices to derive or generate (multiple) DMRS patterns for multi-layer, multi-user, or multi-panel scenarios. Note that the non-uniformity of the derived patterns across layers, users, or panels may cause interference; however, interference can be mitigated using beam-domain multiplexing similar to that done for such cases, and thus performance can be maintained. On the other hand, applying constraints to formulate the derived pattern techniques (such as, grooming and OCC) may be beneficial for multiplexing DMRS for multiple layers, multiple users, and multiple panels if overhead reduction for signaling the patterns is considered.

[0232] The derivation algorithm may be ready to be generally applied to all reference signals for overhead reduction purposes, e.g., CSI-RS, and it may be considered (e.g., jointly) with CSI feedback compression.

[0233] In a method (or apparatus) according to any aspect, the UE may, e.g., report its ability to derive patterns from the channel and (suggested) patterns, and the gNB may, e.g., configure pattern derivation parameters and a selected set of DMRS patterns for the UE. Additionally, for flexible pattern scheduling, the gNB may indicate the selected DMRS pattern for DL transmission to the UE in the DCI, and optionally, the UE may report / indicate its preferred DMRS pattern for the next transmission in the UL UCI.

[0234] In this specification, any presented connection in the described embodiments should be understood in a manner that the components involved are operably coupled. Thus, the connection may be direct or indirect, with any number or combination of intermediate elements, and there may only be a functional relationship between the components.

[0235] In addition, any method, process, and action described or illustrated herein can be implemented using executable instructions in a general or special-purpose processor and stored on a computer-readable storage medium (e.g., disk, memory, etc.) for execution by such a processor. The reference to "computer-readable storage medium" should be understood to cover dedicated circuits such as FPGAs, ASICs, signal processing devices, and other devices.

[0236] The expression "A and / or B" is considered to include any one of the following three scenarios: (i) A, (ii) B, (iii) A and B. In addition, the article "a" should not be understood as "one", that is, the use of the expression "a component" does not exclude the existence of other components. The term "comprising" should be understood in an open sense, that is, in such a way that an object "comprising component A" may also include other components in addition to component A. In addition, the term "comprising" may be limited to "consisting of", that is, consisting only of the specified components.

[0237] Expressions such as "at least one of the following: "<list of two or more components>" and "at least one of <list of two or more components>" and similar phrasings, where the list of two or more components is joined by "and" or "or", mean at least any one component, or at least any two or more components, or at least all components.

[0238] It should be understood that all presented embodiments are merely exemplary, and any feature presented for a particular exemplary embodiment can be used alone with any aspect, or in combination with any feature presented for the same or another particular exemplary embodiment, and / or in combination with any other feature not mentioned. In particular, the exemplary embodiments presented in this specification should also be understood to be disclosed in all possible combinations with each other, as long as it is technically reasonable and the exemplary embodiments are not alternatives to each other. It will also be understood that any feature presented for an exemplary embodiment in a particular category (method / apparatus / computer program / system) can also be used in a corresponding manner in any other category of exemplary embodiments. It should also be understood that the presence of a feature in the presented exemplary embodiments does not necessarily mean that the feature forms an essential feature and cannot be omitted or replaced.

[0239] The statement that a feature includes at least one of the subsequently listed features is not mandatory, that is, the feature includes all of the subsequently listed features or at least one of the multiple subsequently listed features. In addition, it is possible to select the listed features in any combination or only select one of the listed features. A particular combination of all of the subsequently listed features can also be considered. In addition, it is also possible to have only one of the multiple listed features.

[0240] The order of all the method steps presented above is not mandatory, and alternative orders are possible. However, the specific method step sequence shown in the figures should be regarded as a possible method step sequence for each of the embodiments described by the respective figures.

[0241] The subject matter has been described above by way of example embodiments. It should be noted that there are alternative ways and variations that will be apparent to those skilled in the art and that can be implemented without departing from the scope of the appended claims.

[0242] List of abbreviations:

[0243] 3GPP Third Generation Partnership Project

[0244] 5G Fifth Generation

[0245] AI Artificial Intelligence

[0246] BER Bit Error Rate

[0247] CNN Convolutional Neural Network

[0248] CSI Channel State Information

[0249] CSI-RS Channel State Information Reference Signal

[0250] CQI Channel Quality Indicator

[0251] CRI CSI-RS Resource Indicator

[0252] CNN Convolutional Neural Network

[0253] DCI Downlink Control Information

[0254] DMRS Demodulation Reference Signal

[0255] DL Downlink

[0256] FDD Frequency Division Duplexing

[0257] gNB Next Generation Node B

[0258] IE Information Element

[0259] L1-RSRP Layer 1 RSRP

[0260] MAC CE MAC Control Element

[0261] MCS Modulation and Coding Scheme

[0262] ML Machine Learning

[0263] MSE Mean Square Error

[0264] MU Multi-User

[0265] NN Neural Network

[0266] NR New Radio

[0267] PMI Precoding Matrix Indicator

[0268] RAN Radio Access Network

[0269] RB Residual Block

[0270] RRC Radio Resource Control

[0271] RS Reference Signal

[0272] RSRQ Reference Signal Received Quality

[0273] RSRP Reference Signal Received Power

[0274] RSSI Received Signal Strength Indicator

[0275] SU Single User

[0276] SRS Sounding Reference Signal

[0277] UCI Uplink Control Indication

[0278] UE User Equipment

[0279] UL Uplink

Claims

1. A first device, comprising: - means for obtaining channel measurements of a radio channel between the first device and a second device measured on a resource grid; - means for generating a demodulation reference signal pattern on the resource grid based on the channel measurements; and - means for transmitting the generated demodulation reference signal pattern to the second device.

2. The first device according to claim 1, wherein the demodulation reference signal pattern is generated by using a trained machine learning algorithm fed with the channel measurements.

3. The first device according to claim 1 or 2, wherein the first device further comprises: - means for receiving from the second device at least one of the following: - a first information element indicating a maximum number of resource elements in the resource grid to be used for insertion of the demodulation reference signal; and - a second information element indicating channel recovery performance associated with the demodulation reference signal pattern, and wherein the demodulation reference signal pattern is further generated based on at least one of the first information element and the second information element.

4. The first device according to any one of claims 1 to 3, wherein the first device further comprises: - means for generating a plurality of demodulation reference signal patterns on the resource grid based on the channel measurements; and - means for transmitting the plurality of demodulation reference signal patterns to the second device, wherein the plurality of demodulation reference signal patterns correspond to different maximum numbers of resource elements in the resource grid to be used for insertion of the demodulation reference signal, or correspond to different channel recovery performances associated with the respective demodulation reference signal patterns.

5. The first device according to claim 4, wherein the first device further comprises: - means for transmitting a third information element to the second device, the third information element indicating the channel recovery performance associated with the respective demodulation reference signal pattern among the plurality of demodulation reference signal patterns.

6. The first device according to claim 4 or 5, wherein the first device further comprises: - means for transmitting the plurality of demodulation reference signal patterns as nested demodulation reference signal patterns.

7. The first device according to any one of claims 1 to 6, wherein the channel measurements are performed based on: a predetermined demodulation reference signal pattern, and successful decoding of data transmitted on the resource grid together with a demodulation reference signal inserted after the predetermined demodulation reference signal pattern.

8. The first device according to any one of claims 1 to 7, wherein the first device further comprises: - means for transmitting a fourth information element to the second device, the fourth information element indicating the ability of the first device to generate a demodulation reference signal pattern customized for the radio channel.

9. The first device according to any one of claims 1 to 8, wherein the first device further comprises: - means for detecting a change in the radio channel; - means for triggering a fallback to a predetermined demodulation reference signal pattern; - A component for generating a new demodulation reference signal pattern on the resource grid based on new channel measurements performed after the detection of the change in the radio channel, the new channel measurements being based on the predetermined demodulation reference signal pattern.

10. The first apparatus according to any one of claims 1 to 9, wherein the first apparatus further comprises: - A component for sending one or more past demodulation reference signal patterns to the second apparatus, the one or more past demodulation reference signal patterns being retrieved from the memory of the first apparatus and customized for the radio channel during a past communication session between the first apparatus and the second apparatus.

11. A second apparatus, comprises: - A component for receiving a demodulation reference signal pattern from a first apparatus on a resource grid, wherein the demodulation reference signal pattern is generated based on channel measurements of the radio channel between the first apparatus and the second apparatus measured on the resource grid.

12. The second apparatus according to claim 11, wherein the second apparatus further comprises: - A component for receiving a plurality of demodulation reference signal patterns from the first apparatus, wherein the plurality of demodulation reference signal patterns correspond to different maximum numbers of resource elements in the resource grid to be used for the insertion of demodulation reference signals, or correspond to different channel recovery performances associated with the respective demodulation reference signal patterns.

13. The second apparatus according to claim 12, wherein the second apparatus further comprises: - A component for receiving a third information element from the first apparatus, the third information element indicating the channel recovery performance associated with the respective demodulation reference signal pattern among the plurality of demodulation reference signal patterns.

14. The second apparatus according to claim 12 or 13, wherein the second apparatus further comprises: - A component for selecting a demodulation reference signal pattern from the plurality of demodulation reference signal patterns to achieve a given error performance metric, wherein the error performance metric is determined based on at least one of the following: - The effective number of resource elements used for the insertion of demodulation reference signals in the selected demodulation reference signal pattern; - The channel recovery performance for the selected demodulation reference signal pattern; and - The channel measurements of the radio channel.

15. The second apparatus according to any one of claims 12 to 14, wherein the second apparatus further comprises: - A component for receiving the plurality of demodulation reference signal patterns as nested demodulation reference signal patterns.

16. The second apparatus according to any one of claims 11 to 15, wherein the second apparatus further comprises: - A component for sending to the first apparatus at least one of the following: - A first information element indicating the maximum number of resource elements in the resource grid to be used for the insertion of demodulation reference signals; and - A second information element indicating the channel recovery performance associated with the demodulation reference signal pattern. and wherein the demodulation reference signal pattern is further generated based on at least one of the first information element and the second information element.

17. The second apparatus according to any one of claims 11 to 16, wherein the second apparatus further comprises: - means for receiving a fourth information element from the first apparatus, the fourth information element indicating the ability of the first apparatus to generate a demodulation reference signal pattern customized for the radio channel.

18. The second apparatus according to any one of claims 11 to 17, wherein the second apparatus further comprises: - means for receiving one or more past demodulation reference signal patterns from the first apparatus, the one or more past demodulation reference signal patterns being retrieved from the memory of the first apparatus and customized for the radio channel during a past communication session between the first apparatus and the second apparatus.

19. A first method, comprising: - obtaining channel measurements of a radio channel between a first apparatus and a second apparatus measured on a resource grid; - generating a demodulation reference signal pattern on the resource grid based on the channel measurements; and - transmitting the generated demodulation reference signal pattern.

20. The first method according to claim 19, wherein the demodulation reference signal pattern is generated by a trained machine learning algorithm fed with the channel measurements.

21. The first method according to claim 19 or 20, wherein the first method further comprises: - receiving at least one of the following: - a first information element indicating a maximum number of resource elements in the resource grid to be used for insertion of the demodulation reference signal; and - a second information element indicating channel recovery performance associated with the demodulation reference signal pattern, and wherein the demodulation reference signal pattern is further generated based on at least one of the first information element and the second information element.

22. The first method according to any one of claims 19 to 21, wherein the first method further comprises: - generating a plurality of demodulation reference signal patterns on the resource grid based on the channel measurements; and - transmitting the plurality of demodulation reference signal patterns, wherein the plurality of demodulation reference signal patterns correspond to different maximum numbers of resource elements in the resource grid to be used for insertion of the demodulation reference signal, or correspond to different channel recovery performances associated with the respective demodulation reference signal patterns.

23. The first method according to claim 22, wherein the first method further comprises: - transmitting a third information element indicating the channel recovery performance associated with the respective demodulation reference signal pattern among the plurality of demodulation reference signal patterns.

24. The first method according to claim 22 or 23, wherein the first method further comprises: - transmitting the plurality of demodulation reference signal patterns as nested demodulation reference signal patterns.

25. The first method according to any one of claims 19 to 24, wherein the channel measurement is performed based on: a predetermined demodulation reference signal pattern, and successful decoding of data transmitted on the resource grid together with the demodulation reference signal inserted after the predetermined demodulation reference signal pattern.

26. The first method according to any one of claims 19 to 25, wherein the first method further comprises: - Transmitting a fourth information element indicating the ability of the first device to generate a demodulation reference signal pattern customized for the radio channel.

27. The first method according to any one of claims 19 to 26, wherein the first method further comprises: - Detecting a change in the radio channel; - Triggering a fallback to a predetermined demodulation reference signal pattern; - Generating a new demodulation reference signal pattern on the resource grid based on new channel measurements performed after the detection of the change in the radio channel, the new channel measurements being based on the predetermined demodulation reference signal pattern.

28. The first method according to any one of claims 19 to 27, wherein the first method further comprises: - Transmitting one or more past demodulation reference signal patterns retrieved from the memory of the first device and customized for the radio channel during a past communication session between the first device and the second device.

29. A second method, comprising: - Receiving a demodulation reference signal pattern on the resource grid, wherein the demodulation reference signal pattern is generated based on channel measurements of the radio channel between the first device and the second device measured on the resource grid.

30. The second method according to claim 29, wherein the second method further comprises: - Receiving a plurality of demodulation reference signal patterns, wherein the plurality of demodulation reference signal patterns correspond to different maximum numbers of resource elements in the resource grid that will be used for the insertion of demodulation reference signals, or correspond to different channel recovery performances associated with the respective demodulation reference signal patterns.

31. The second method according to claim 30, wherein the second method further comprises: - Receiving a third information element indicating the channel recovery performance associated with the respective demodulation reference signal pattern among the plurality of demodulation reference signal patterns.

32. The second method according to claim 30 or 31, wherein the second method further comprises: - Selecting a demodulation reference signal pattern from the plurality of demodulation reference signal patterns to achieve a given error performance metric, wherein the error performance metric is determined based on at least one of the following: - The effective number of resource elements used for inserting the demodulation reference signal in the selected demodulation reference signal pattern; - The channel recovery performance for the selected demodulation reference signal pattern; and - The channel measurement of the radio channel.

33. The second method according to any one of claims 30 to 32, wherein the second method further comprises: - Receive the plurality of demodulation reference signal patterns as nested demodulation reference signal patterns.

34. The second method according to any one of claims 29 to 33, wherein the second method further comprises: - Transmit at least one of the following: - A first information element indicating a maximum number of resource elements in the resource grid to be used for inserting demodulation reference signals; and - A second information element indicating a channel recovery performance associated with the demodulation reference signal pattern, and wherein the demodulation reference signal pattern is further generated based on at least one of the first information element and the second information element.

35. The second method according to any one of claims 29 to 34, wherein the second method further comprises: - Receive a fourth information element indicating the ability of the first device to generate a demodulation reference signal pattern customized for the radio channel.

36. The second method according to any one of claims 29 to 35, wherein the second method further comprises: - Receive one or more past demodulation reference signal patterns retrieved from a memory of the first device and customized for the radio channel during a past communication session between the first device and the second device.

37. A system comprising a mobile entity or a part thereof and a server or a part thereof that together perform the method according to any one of claims 19 to 36.

38. A computer program product that, when executed by a processor of a device, causes the device to perform the method according to any one of claims 19 to 36.

39. A computer program that, when executed by a processor, causes a device to perform or control the actions of the method according to any one of claims 19 to 36.

40. A tangible and non-transitory storage medium comprising the computer program product according to claim 38.

Images