Signaling for communicating with learned constellations and sparse demodulation reference signal

Abstract

Systems, methods, apparatuses, and computer program products for communicating with learned constellations and sparse demodulation reference signal (DMRS). A method may include receiving symbols from a network element. The method may also include receiving an array including a number of subcarriers, a number of the symbols, and a number of input channels. The input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The method may further include processing the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the method may include jointly training the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

Description

TITLE:SIGNALING FOR COMMUNICATING WITH LEARNED CONSTELLATIONS AND SPARSE DEMODULATION REFERENCE SIGNALFIELD:

[0001] Some example embodiments may generally relate to mobile or wireless telecommunication systems, such as Long Term Evolution (LTE) or fifth generation (5G) new radio (NR) access technology, or 5G beyond, or sixth generation (6G) access technology, or other communications systems. For example, certain example embodiments may relate to apparatuses, systems, and / or methods for signaling for communicating with learned constellations and sparse demodulation reference signal (DMRS).BACKGROUND:

[0002] Examples of mobile or wireless telecommunication systems may include the Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access Network (UTRAN), Long Term Evolution (LTE) Evolved UTRAN (E-UTRAN), LTE- Advanced (LTE-A), MulteFire, LTE-A Pro, fifth generation (5G) radio access technology or new radio (NR) access technology and / or sixth generation (6G) radio access technology. 5G and 6G wireless systems refer to the next generation (NG) of radio systems and network architecture. 5G and 6G network technology is mostly based on NR technology, but the 5G / 6G (or NG) network can also build on E-UTRAN radio. It is estimated that NR may provide bitrates on the order of 10-20 Gbit / s or higher, and may support at least enhanced mobile broadband (eMBB) and ultra-reliable low-latency communication (URLLC) as well as massive machine-type communication (mMTC). NR is expected to deliver extreme broadband and ultra-robust, low-latency connectivity and massive networking to support the Internet of Things (IoT).SUMMARY:

[0003] Some example embodiments may be directed to a method. The method may includereceiving symbols from a network element. The method may also include receiving an array including a number of subcarriers, a number of the symbols, and a number of input channels, wherein the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The method may further include processing the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the method may include jointly training the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

[0004] Other example embodiments may be directed to an apparatus. The apparatus may include at least one processor and at least one memory including computer program code. The at least one memory and the computer program code may be configured to, with the at least one processor, cause the apparatus at least to receive symbols from a network element. The apparatus may also be configured to receive an array comprising a number of subcarriers, a number of the symbols, and a number of input channels, wherein the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The apparatus may further be configured to process the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the apparatus may be configured to jointly train the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

[0005] Other example embodiments may be directed to an apparatus. The apparatus may include means for receiving symbols from a network element. The apparatus may also include means for receiving an array including a number of subcarriers, a number of the symbols, and a number of input channels, wherein the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The apparatus may further include means for processing the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the apparatus may include means for jointly training the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

[0006] In accordance with other example embodiments, a non-transitory computer readable medium may be encoded with instructions that may, when executed in hardware, perform a method. The method may include receiving symbols from a network element. The method may also include receiving an array including a number of subcarriers, a number of the symbols, and a number of input channels, wherein the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The method may further include processing the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the method may include jointly training the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

[0007] Other example embodiments may be directed to a computer program product that performs a method. The method may include receiving symbols from a network element. The method may also include receiving an array including a number of subcarriers, a number of the symbols, and a number of input channels, wherein the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The method may further include processing the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the method may include jointly training the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

[0008] Other example embodiments may be directed to an apparatus that may include circuitry configured to receive symbols from a network element. The apparatus may also include circuitry configured to receive an array including a number of subcarriers, a number of the symbols, and a number of input channels, wherein the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The apparatus may further include circuitry configured to process the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the apparatus may be configured to jointly train the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

[0009] Further example embodiments may be directed to a method. The method may include receiving, from a user equipment, a channel quality indicator. The method may also include selecting, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. The method may further include adjusting the modulation coding scheme and a demodulation reference signal density based on a block error rate.

[0010] Other example embodiments may be directed to an apparatus. The apparatus may include at least one processor and at least one memory including computer program code. The at least one memory and the computer program code may be configured to, with the at least one processor, cause the apparatus at least to receive, from a user equipment, a channel quality indicator. The apparatus may also be configured to select, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. The apparatus may further be configured to adjust the modulation coding scheme and a demodulation reference signal density based on a block error rate.

[0011] Other example embodiments may be directed to an apparatus. The apparatus may include means for receiving, from a user equipment, a channel quality indicator. The apparatus may also include means for selecting, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. The apparatus may further include means for adjusting the modulation coding scheme and a demodulation reference signal density based on a block error rate.

[0012] In accordance with other example embodiments, a non-transitory computer readable medium may be encoded with instructions that may, when executed in hardware, perform a method. The method may include receiving, from a user equipment, a channel quality indicator. The method may also include selecting, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. The method may further include adjusting the modulation coding scheme and a demodulation reference signal density based on a block error rate.

[0013] Other example embodiments may be directed to a computer program product that performs a method. The method may include receiving, from a user equipment, a channelquality indicator. The method may also include selecting, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. The method may further include adjusting the modulation coding scheme and a demodulation reference signal density based on a block error rate.

[0014] Other example embodiments may be directed to an apparatus that may include circuitry configured to receive, from a user equipment, a channel quality indicator. The apparatus may also include circuitry configured to select, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. The apparatus may further include circuitry configured to adjust the modulation coding scheme and a demodulation reference signal density based on a block error rate.BRIEF DESCRIPTION OF THE DRAWINGS:

[0015] For proper understanding of example embodiments, reference should be made to the accompanying drawings, wherein:

[0016] FIG. 1 illustrates an example approach for transmission of two multiple input multiple output (MIMO) layers, according to certain example embodiments.

[0017] FIG. 2 illustrates an example single input-single output (SISO) DeepRx model, according to certain example embodiments.

[0018] FIG. 3 illustrates an example DeepRx MIMO extension, according to certain example embodiments.

[0019] FIG. 4 illustrates an example supervised training of the constellation shape(s), according to certain example embodiments.

[0020] FIG. 5 illustrates an example signal diagram, according to certain example embodiments.

[0021] FIG. 6 illustrates an example outer loop link adaptation for adjusting a modulation coding scheme (MCS) and demodulation reference signal (DMRS) density in an uplink (UL) scenario, according to certain example embodiments.

[0022] FIG. 7 illustrates example DMRS configurations, according to certain example embodiments.

[0023] FIG. 8 illustrates an example resulting block error rate (BLER) performance, according to certain example embodiments.

[0024] FIG. 9 illustrates an example flow diagram of a method, according to certain example embodiments.

[0025] FIG. 10 illustrates an example flow diagram of another method, according to certain example embodiments.

[0026] FIG. 11 illustrates a set of apparatuses, according to certain example embodiments.DETAILED DESCRIPTION:

[0027] It will be readily understood that the components of certain example embodiments, as generally described and illustrated in the figures herein, may be arranged and designed in a wide variety of different configurations. The following is a detailed description of some example embodiments of systems, methods, apparatuses, and computer program products for communicating with learned constellations and sparse demodulation reference signal (DMRS). In certain example embodiments, the signaling may be performed with machine learning (ML) receivers.

[0028] The features, structures, or characteristics of example embodiments described throughout this specification may be combined in any suitable manner in one or more example embodiments. For example, the usage of the phrases “certain embodiments,” “an example embodiment,” “some embodiments,” or other similar language, throughout this specification refers to the fact that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment. Thus, appearances of the phrases “in certain embodiments,” “an example embodiment,” “in some embodiments,” “in other embodiments,” or other similar language, throughout this specification do not necessarily refer to the same group of embodiments, and the described features, structures, or characteristics may be combined in any suitable manner in one or more example embodiments. Further, the terms “base station”, “cell”, “node”, “gNB”, “network” or other similar language throughout this specification may be used interchangeably.

[0029] As used herein, “at least one of the following: ” and “at least one of ” and similar wording, where the list of two or more elements are joined by “and” or “or,” mean at least any one of the elements, or at least any two or more of the elements, or at least all the elements.

[0030] Use of learned constellations for pilotless transmissions and implementation of ML- based have been proposed. Signaling of learned constellations and demodulation reference signal (DMRS) density reduction have also been performed. However, there are no works for combining conventional pilots (or DMRS configurations) with learned constellations in the context of multiple input multiple output (MIMO) systems. Additionally, there is currently no way to seamlessly select between pilotless constellation learning and pilotbased detection. In certain cases, training a completely pilotless link may not be the best approach in terms of complexity and obtained throughput gain, especially in high-order MIMO systems where an excessively large machine learning (ML)-based receiver model(from here on referred to as “DeepRx”) might be needed to support pilotless detection. Consequently, there is a need for a solution that can harness the benefits of pilotbased channel estimation and constellation shaping to achieve a more optimized and balanced air interface.

[0031] In view of the drawbacks described above, certain example embodiments may be related to artificial intelligence (Al)-native air interface, where the physical layer is built around ML-based algorithms. This may involve ML both at the transmitter and at the receiver. At the receiver side, a fully learned receiver may be utilized, such as DeepRx. At the transmitter side, the waveform may be optimized for DeepRx by training the Tx-Rx link end-to-end. An example of such optimization may include learning pilotless transmissions, which may be possible by jointly learning the transmitter constellation shape and the DeepRx receiver weights. However, since learning such a pilotless scheme may be challenging in MIMO systems where spatial multiplexing is also performed, a large model or ways of trading off complexity with performance may be needed.

[0032] FIG. 1 illustrates an example approach for transmission of two MIMO layers, according to certain example embodiments. As illustrated in FIG. 1, the learnedconstellation of the transmitters 100 and the DeepRx of the receivers 105 represent elements that are being learned. In certain example embodiments, the learned constellation and using sparse DMRS may be combined, as well as feeding a DMRS density parameter to the DeepRx model (for supporting varying DMRS densities). A reduced DMRS density will increase the throughput as higher proportion of the available resources can be used for useful data transmission.

[0033] According to certain example embodiments, an air interface may be provided where pilot-based detection is combined with constellation shaping and fully learned receivers. This approach may reduce the complexity of the ML-based receiver by not resorting to fully pilotless transmissions, while still taking advantage of the benefits of learned constellations in the form of reduced DMRS overhead.

[0034] To achieve the various advantageous effects of certain example embodiments described herein, a transmit signal type may be defined where conventional DMRS is combined with learned constellation shapes, either separate or shared between different MIMO layers. Additionally, sparse DMRS configurations may be defined to achieve throughput gains with learned constellations optimized for sparse DMRS. As an implementation aspect, the DMRS density may also be provided as an additional input to the DeepRx model for supporting different DMRS densities with a single DeepRx model. Furthermore, in some example embodiments, a signaling procedure for agreeing on the learned constellation parameters and used DMRS density may be defined. The learned constellation parameters and the used DMRS may be dependent parameters. The signaling procedure may be for communicating the selected constellation library and DMRS density for physical uplink shared channel (PUSCH) transmissions. Other example embodiments may introduce a mechanism for integrating the density of DMRS into a link adaptation algorithm and associated control signaling. For instance, signaling may be for an extended outer loop link adaptation (OLLA), and the link adaptation may be for adjusting the modulation and coding scheme (MCS) and DMRS density.

[0035] FIG. 2 illustrates an example single input-single output (SISO) DeepRx model, according to certain example embodiments. As illustrated in FIG. 2, the DeepRx modelmay detect a signal with sparse DMRS and learned constellation. The frequency-domain orthogonal frequency-division multiplexing (OFDM) symbols may be fed to a Residual Network (ResNet) type convolutional neural network (CNN), whose input may be one time transmission interval (TTI). The TTI may include 14 OFDM symbols (denoted as Nsymb).

[0036] As illustrated in FIG. 2, the number of subcarriers (SCs) fed to the receiver may be denoted by Nsc. That is, Nscis the number of received SCs on the active BWP, Nsymb is the number of OFDM symbols in a slot (e.g., 14), Ni ... Nj+k+i denote the numbers of output channels in the convolutional layers inside the individual ResNet blocks, and NB denotes the number of bits per resource element (RE). Altogether, the ML receiver may be fed a real -valued Nscx Nsymb x 3 array, where the last dimension represents the number of input channels. The input channels may include a received signal, a raw channel estimate, and a DMRS density scalar value that may be broadcast across the whole resource grid. According to certain example embodiments, in the case of multiple input multiple output (MIMO), the channel matrix may be vectorized along the channel dimension, resulting in NLNR raw channel estimates per RE, where NL is the number of layers, NR is the number of Rx antennas (correspondingly, the received OFDM symbol array may include NR samples per RE), and Ni ... Nj+k+i denotes the numbers of output channels in the convolutional layers inside the ResNet blocks. This array may be processed with one or more ResNet blocks. In some example embodiments, if the number of channels between two consecutive ResNet blocks is the same, the convolutional layer of the skip connection may be omitted.

[0037] As illustrated in FIG. 2, block 200 refers to repeating the denoted block M times. Block 205 refers to the input for the received signal, and block 210 refers to the input for the raw channel estimate. Additionally, block 215 refers to the input for used DMRS density, which is replicated for each RE. Block 220 refers to the transforming complexvalued signals to real-valued by concatenating the real and imaginary parts. Block 225 refers to the first convolutional layer, and block 230 refers to the second convolutional layer. Further, block 235 refers to the third convolutional layer, and block 240 refers to refers to the summation operation between the two branches. Block 245 refers to the fourthconvolutional layer, and block 250 refers to the output node consisting of LLRs.

[0038] According to certain example embodiments, the output of the ResNet blocks may be fed to a final 2-dimensional convolutional layer, at the output of which the final loglikelihood ratio (LLR) estimates may be obtained. The LLR estimates may be the receiver’s estimates of the probabilities of the transmitted bits. That is, the original transmitted bits may represent the ideal output of the model. In some example embodiments, the DeepRx may be trained with simulated data, where access to the original transmit sequence is not an issue. Extension to MIMO may be done by utilizing the similar input with DeepRx models tailored for MIMO detection.

[0039] FIG. 3 illustrates an example DeepRx MIMO extension, according to certain example embodiments. In particular, FIG. 3 illustrates a DeepRx MIMO extension where Min denotes the number of output channels from the PreDeepRx, while Mexdenotes the number of channels after sparse expansion. As illustrated in FIG. 3, the PreDeepRx and DeepRx components follow the architecture of FIG. 2, while the blocks in between them represent a learned multiplication between the input channels of the ML receiver. In particular, the sparse expansion represents multiplication of the input channels by a learned sparse matrix, after which there is a learned multiplier for the imaginary part of each of the expanded channels. The number of expanded channels, denoted by Mex, may be divisible by 3, as the channels are then split into three parts. Two of these parts are multiplied element wise and concatenated with the third part. Finally, the resulting channels are fed to the DeepRx part of the model, which is similar to the architecture shown in FIG. 2.

[0040] In both of these cases illustrated in FIGs. 2 and 3, the DeepRx model may be trained in a supervised manner (i.e., random data is generated by saving the original transmitted bits and the received data samples after the receiver FFT). The latter may be the input of the ML model, while the former may be the desired output. Binary cross entropy can be used as the loss function, based on which the parameters of the models are updated using stochastic gradient descent. In some example embodiments, adaptive moment estimation (ADAM) can be used as the stochastic gradient descent method.

[0041] In certain example embodiments, the constellation shape of the transmitter may belearned under a sparse DMRS configuration. For instance, a single constellation may be learned for a range of densities under different channel conditions. The learned approach may refer to the transmitter being able to associate a constellation shape with a DMRS density parameter. Another approach may include learning a single constellation for several DMRS densities. The learning may mean that the constellation shape is optimized for a given DMRS density(ies). In other example embodiments, it may also be possible to learn a different constellation shape for each DMRS density in which case the DMRS density may be mapped to a certain constellation.

[0042] According to further example embodiments, a separate constellation may be needed for each modulation order (e.g., 2, 4, 6, 8 bits per symbol, etc.). The different modulation orders may be detected either with separately trained DeepRx models, or with a single model trained to support all constellation sizes. However, in other example embodiments, a single DeepRx model may be used for most scenarios in an effort to reduce the number of separate models (particularly if there are also different constellations for each DMRS). The DMRS density input parameter may provide an easier means to train a single DeepRx model for varying DMRS entities.

[0043] In certain example embodiments, the constellation shape(s) may be learned with the procedure illustrated in FIG. 4. In particular, FIG. 4 illustrates an example supervised training of the constellation shape(s), according to certain example embodiments. As illustrated in FIG. 4, the input (e.g., random bits 400 and / or random neural network (NN) model parameters 415) may be considered to be encoded bits, while the output (e.g., trained neural network (NN) model parameters 415) may be the estimated log-likelihood ratios (LLRs), interpreted as bit probabilities. The non-leamed elements may be included as fixed layers, and these fixed layers may include all conventional transmitter and receiver functionalities, as well as a randomized channel model. Additionally, binary cross entropy may be used as the loss function. In model forward pass (405) the end-to-end model with the current model parameters is inferenced and the loss function is calculated using the predefined loss function, e.g., binary cross entropy (410). During model backward pass (425), the model parameters are updated based on stochastic gradient descent (SGD). Oncea predefined number of backward passes have been completed, the learned model parameters are provided as the output of the learning procedure (415).

[0044] According to certain example embodiments, the practical use of a waveform with a learned constellation and sparse DRS pattern may be standardized. For instance, the learned constellations may be predefined and stored in a look-up table (LUT) on each device (e.g., BS and UEs). To support all modulation orders, there may be a library of learned constellations for each modulation order (e.g., 2, 4, 6, and 8 bits per symbol). Additionally, the DMRS density may be defined as a density parameter which impacts how many DMRS symbols are discarded from the original pattern. In some example embodiments, the constellation shape may be dependent upon the DMRS density, which means that the selected constellation library and DMRS density are dependent parameters. According to certain example embodiments this dependency may mean that a particular constellation shape supports only certain DMRS densities, or that each DMRS density is associated with a different constellation shape.

[0045] FIG. 5 illustrates an example signal diagram, according to certain example embodiments. For instance, the signal diagram may be for communicating the selected constellation library and DMRS density for PUSCH. As illustrated in FIG. 5, at 510, a connection setup may be established between the BS 500 and the UE 505. At 515, the UE 505 indicates to the BS 500 the UE’s 505 capability of using learned constellations with sparse DMRS during the connection setup. Once the UE 505 has indicated its capability to the BS 500, the BS 500 may decide to initiate more efficient PUSCH data transfers using a particular learned constellation library with a combined DMRS density parameter. Thus, at 520, the BS 500 indicates the constellation library and the DMRS density parameter to the UE 505. After signaling the constellation library and the DMRS density to the UE 505, it may be assumed that the PUSCH data is transmitted with this configuration. It is also possible to revert back to regular PUSCH with conventional quadrature amplitude modulation (QAM) constellation and full DMRS. Additionally, it is possible to finetune the DMRS density, either by using the same constellation if supported, or by changing the constellation based on the DMRS density.

[0046] At 525, the UE 505 transmits a scheduling request (physical uplink control channel (PUCCH)) to the BS 500, and in response, at 530, the BS 500 transmits a scheduling grant (physical downlink control channel (PDCCH)) to the UE 505. At 530, the UE 505 may also activate the proper constellation library and use reduced DMRS with the defined DMRS density. The activation of the proper constellation library and use of the reduced DMRS with the defined density may occur once the UE 505 receives the scheduling grant (PDCCH) from the BS 500. At 535, the UE 505 transmits data with the learned constellation (PUSCH). At 540, the BS 500 deactivates the learned constellation and DMRS density reduction, and at 545, the UE 505 reverts to regular PUSCH data transmission with regular QAM and full DMRS.

[0047] In certain example embodiments, as DeepRx may obtain the initial channel estimates as input alongside the received data, where the given channel estimates are sufficiently accurate. Although full DMRSs may result in accurate channel estimates, they may also create excessive pilot overhead. On the other hand, if DMRS configuration is sparse, the initial channel estimates may have too much error for the DeepRx to extract useful information from them. Thus, it may be desirable to determine an optimal number (e.g., optimal number of initial channel estimates / optimal number of DMRS symbols) and an optimal configuration of DMRS symbols that result in the best performance when combined with a jointly learned constellation. Furthermore, the configuration of the pilots (i.e., on which OFDM symbol indices and subcarriers they are allocated) may influence the accuracy of the initial channel estimates (e.g., raw channel estimates calculated from DMRS where least squares estimation may be used) and, thus, the overall throughput performance of the system.

[0048] According to certain example embodiments, there may be multiple ways to implement a varying DMRS density. For instance, in one example embodiment, the density may be defined as the proportion of pilot symbols retained with respect to the full DMRS pattern. For example, if the DMRS density is set to 75%, every fourth DMRS symbol is omitted. Alternatively, DMRS symbols may be removed in a predefined manner in the frequency or time directions of the resource grid. In these cases, it may be enoughto signal the density parameter to the transmitter as that would completely define the DMRS pattern.

[0049] In other example embodiments, different predefined DMRS patterns may be used for different DMRS density values. The complete DMRS pattern may either be signaled or be a priori known to the UE. Additionally, in this case, no density parameter would be needed as the DMRS pattern would determine it directly. Regardless of the specific approach, it may be assumed that signaling a DMRS density parameter may define the DMRS configuration, or alternatively, signaling a DMRS configuration may define the DMRS density. Additionally, in some example embodiments, each DMRS density may be associated with a predefined constellation shape (either a single constellation that supports varying densities, or a different predefined constellation for each density).

[0050] According to certain example embodiments, the UE may calculate a channel quality indicator (CQI) and report it to the BS. The BS may then select an initial MCS for the UE. In some example embodiments, an outer loop link adaptation (OLLA) algorithm may monitor the block error rate (BLER), and adjust the MCS for each packet transmission so that a predefined target error rate (e.g., 10%) is satisfied. The BS may be the receiving end, and the BS may not need CQI or error reporting from the UE.

[0051] In certain example embodiments, adjusting the DMRS density may serve as an additional optimization step in OLLA. The link adaptation may involve selecting a learned constellation, coding rate, and DMRS density level (or configuration). Selecting the optimal combination of the learned constellation, coding rate, and DMRS density level may be accomplished via an optimization algorithm or using ML including, for example, reinforcement learning.

[0052] FIG. 6 illustrates an example outer loop link adaptation for adjusting the MCS and DMRS density in an uplink (UL) scenario, according to certain example embodiments. In particular, FIG. 6 illustrates a BS 615 with an adjusted OLLA 625, where an optimization algorithm or ML-based model takes as input cyclic redundancy check (CRC) status, CQI, channel statistics, and possible other information 620 and determines the MCS (using a learned constellation) and DMRS density based on a target error rate 630 andcommunicates it to the UE 610. According to certain example embodiments, there may be several ways to implement the optimization in OLLA. Lor instance, in one example embodiment, the DMRS density may be incorporated to OLLA using an MCS table with learned constellations, where each MCS index may have a certain number of different DMRS density levels. In another example embodiment, the BS may decide the MCS based on a conventional procedure (e.g., OLLA) using full DMRS. After establishing the MCS, the DMRS density may be reduced. In further example embodiments, the BS may use a reinforcement learning algorithm to choose an optimal MCS and DMRS density based on different inputs such as the reported CQI, reported errors, traffic type, or channel conditions. In this case, the RL agent would output the MCS and DMRS density, based on the input state consisting of CQI, reported errors, traffic type, or channel conditions. The reward would be determined in terms of the achieved throughput, guiding the RL agent in determining the MCS and DMRS densities maximizing the throughput.

[0053] According to certain example embodiments, the DMRS densities and MCS may be communicated to the UE by the BS. In certain example embodiments, the OLLA may not require any new feedback from the UE, but may determine both the DMRS density and the MCS. Assuming uplink (UL), OLLA may be provided with the channel information (e.g., reported input such as CQI, errors, traffic type, channel conditions, additional statistics such as estimated Doppler, etc.) from the UE to determine the correct DMRS density.

[0054] Validation of the benefits of the signaling for communicating with learned constellations and reduced pilots with ML receivers may be performed by training two different DeepRx models with sparse DMRS configurations for a system with two MIMO layers. Lor instance, one DeepRx model may include (learned constellations + DeepRx, sparse DMRS), which may involve the use of a DeepRx relying on DMRS pilots and a learned constellation such that the DMRS density is one third of the conventional baseline system. The constellations under this DeepRx model may be learned separately for the two MIMO layers. Another DeepRx model may include (QAM + DeepRx, sparse DMRS), which may involve DeepRx relying on DMRS pilots and conventional QAM constellation, such that the DMRS density is one third of the conventional baseline system.

[0055] FIG. 7 illustrates example DMRS configurations, according to certain example embodiments. In particular, FIG. 7 illustrates a sparsified DMRS pattern, and leaves out predefined pilots (every 3rd DMRS symbol is left out twice, i.e., the density is 33%). As illustrated in FIG. 7, the sparse DMRS configuration carries symbols in OFDM symbol indices 2 and 11, but the DMRSs are at every 6th subcarrier. For the two layers, an offset of 3 subcarriers may be applied for balancing the sparse DMRS across the used bandwidth. The least squares (LS) channel estimates may be computed from the DMRSs and interpolated linearly over the whole resource grid and given as input to the DeepRx alongside the received signal. The training of DeepRx models may be carried out pilotless and based on conventional DeepRx. Furthermore, the two DeepRx versions may have identical architectures based on the fully learned multiplicative transformation.

[0056] According to certain example embodiments, in addition to the DeepRx models, two baseline methods that use a conventional QAM constellation and a K-Best detector may be provided. For example, one baseline method may involve (QAM + LS + K-Best, full DMRS), where conventional DMRS pilot configuration is applied in which DMRS pilots are in symbol indices 2 and 11 , and every 2nd subcarrier. That is, the whole OFDM symbol may be used for pilots for two players (see FIG. 7), which means that the baseline may use 3x more DMRS pilots than DeepRx. Furthermore, the LS channel estimates may be linearly interpolated over the resource grid. Another baseline method may involve (QAM + K-Best, known channel), where the perfect channel state information is used.

[0057] FIG. 8 illustrates an example resulting BLER performance, according to certain example embodiments. As illustrated in FIG. 8, both DeepRx models outperform the LS- based practical baseline (QAM + LS + K-Best, full DMRS) despite using one third of the amount of DMRSs. When comparing the two different DeepRx models, it can be observed that using a learned constellation significantly reduces the BLER compared to only using a conventional QAM constellation. The simulation result indicates that combining sparse DMRS with learned constellations improves the link throughput.

[0058] FIG. 9 illustrates an example flow diagram of a method, according to certain example embodiments. In an example embodiment, the method of FIG. 9 may beperformed by a network entity, or a group of multiple network elements in a 3 GPP system, such as LTE or 5G-NR. For instance, in an example embodiment, the method of FIG. 9 may be performed by a UE, similar to one of apparatuses 10 or 20 illustrated in FIG. 11.

[0059] As illustrated in FIG. 9, the method may include, at 900, receiving symbols from a network element. The method may also include, at 905, receiving an array including a number of subcarriers, a number of the symbols, and a number of input channels. According to certain example embodiments, the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The method may further include, at 910, processing the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the method may include, at 915, jointly training the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

[0060] According to certain example embodiments, the method may also include outputting a final log-likelihood ratio estimate based on the processed array. According to some example embodiments, the transmitter constellation shape may be trained under a sparse demodulation reference symbol configuration. According to other example embodiments, the transmitter constellation shape may be modified based on encoded bits and bit probabilities.

[0061] In certain example embodiments, the method may further include mapping a demodulation reference signal density to the transmitter constellation shape. In some example embodiments, the demodulation reference signal density may be associated with either a single constellation that supports varying demodulation reference signal densities, or a different predefined constellation for each demodulation reference signal density. In other example embodiments, the method may also include assigning a separate constellation for different modulation orders, and detecting the different modulation orders. In further example embodiments, the detection may be performed with separately trained receiver models, or with a single receiver model trained to support a plurality of constellation sizes.

[0062] FIG. 10 illustrates an example flow diagram of another method, according to certainexample embodiments. In an example embodiment, the method of FIG. 10 may be performed by a network entity, or a group of multiple network elements in a 3 GPP system, such as LTE or 5G-NR. For instance, in an example embodiment, the method of FIG. 10 may be performed by a gNB, BS, or network, similar to one of apparatuses 10 or 20 illustrated in FIG. 11.

[0063] As illustrated in FIG. 10, the method may include, at 1000, receiving, from a user equipment, a channel quality indicator. The method may also include, at 1005, selecting, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. The method may further include, at 1010, adjusting the modulation coding scheme and a demodulation reference signal density based on a block error rate.

[0064] According to certain example embodiments, the method may also include selecting a modified constellation based on the adjustment of the demodulation reference density. According to some example embodiments, the method may further include incorporating the demodulation reference signal density to an outer loop link adaptation algorithm via a modulation coding scheme table comprising modified constellations. According to other example embodiments, the method may also include selecting the modulation coding scheme and the demodulation reference signal density based on a reported channel quality indicator, a reported error, a traffic type, or channel conditions.

[0065] FIG. 11 illustrates a set of apparatuses 10 and 20 according to certain example embodiments. In certain example embodiments, apparatuses 10 and 20 may be elements in a communications network or associated with such a network. For example, apparatus 10 may be a UE or similar device, and apparatus 20 may be a BS, gNB, network, or other similar computing device.

[0066] In some example embodiments, apparatuses 10 and 20 may include one or more processors, one or more computer-readable storage medium (for example, memory, storage, or the like), one or more radio access components (for example, a modem, a transceiver, or the like), and / or a user interface. In some example embodiments, apparatuses 10 and 20 may be configured to operate using one or more radio accesstechnologies, such as GSM, LTE, LTE-A, NR, 5G, 6G, WLAN, WiFi, NB-IoT, Bluetooth, NFC, MulteFire, and / or any other radio access technologies. It should be noted that one of ordinary skill in the art would understand that apparatuses 10 and 20 may include components or features not shown in FIG. 11.

[0067] As illustrated in the example of FIG. 11, apparatuses 10 and 20 may include or be coupled to a processors 12 and 22 for processing information and executing instructions or operations. Processors 12 and 22 may be any type of general or specific purpose processor. In fact, processors 12 and 22 may include one or more of general-purpose computers, special purpose computers, microprocessors, DSPs, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), and processors based on a multicore processor architecture, as examples. While a single processors 12 and 22 is shown in FIG. 11, multiple processors may be utilized according to other example embodiments. For example, it should be understood that, in certain example embodiments, apparatuses 10 and 20 may include two or more processors that may form a multiprocessor system (e.g., in this case processors 12 may represent a multiprocessor) that may support multiprocessing. According to certain example embodiments, the multiprocessor system may be tightly coupled or loosely coupled (e.g., to form a computer cluster).

[0068] Processors 12 and 22 may perform functions associated with the operation of apparatuses 10 and 20 including, as some examples, precoding of antenna gain / phase parameters, encoding and decoding of individual bits forming a communication message, formatting of information, and overall control of the apparatuses 10 and 20, including processes and examples illustrated in FIGs. 1-8.

[0069] Apparatuses 10 and 20 may further include or be coupled to a memories 14 and 24 (internal or external), which may be respectively coupled to processors 12 and 24 for storing information and instructions that may be executed by processors 12 and 24. Memories 14 and 24 may be one or more memories and of any type suitable to the local application environment, and may be implemented using any suitable volatile or nonvolatile data storage technology such as a semiconductor-based memory device, a magnetic memory device and system, an optical memory device and system, fixedmemory, and / or removable memory. For example, memories 14 and 24 can be comprised of any combination of random access memory (RAM), read only memory (ROM), static storage such as a magnetic or optical disk, hard disk drive (HDD), or any other type of non- transitory machine or computer readable media. The instructions stored in memories 14 and 24 may include program instructions or computer program code that, when executed by processors 12 and 22, enable the apparatuses 10 and 20 to perform tasks as described herein.

[0070] In certain example embodiments, apparatuses 10 and 20 may further include or be coupled to (internal or external) a drive or port that is configured to accept and read an external computer readable storage medium, such as an optical disc, USB drive, flash drive, or any other storage medium. For example, the external computer readable storage medium may store a computer program or software for execution by processors 12 and 22 and / or apparatuses 10 and 20 to perform any of the methods and examples illustrated in FIGs. 1- 8.

[0071] In some example embodiments, apparatuses 10 and 20 may also include or be coupled to one or more antennas 15 and 25 for receiving a downlink signal and for transmitting via an UL from apparatuses 10 and 20. Apparatuses 10 and 20 may further include a transceivers 18 and 28 configured to transmit and receive information. The transceivers 18 and 28 may also include a radio interface (e.g., a modem) coupled to the antennas 15 and 25. The radio interface may correspond to a plurality of radio access technologies including one or more of GSM, LTE, LTE-A, 5G, NR, 6G, WLAN, NB-IoT, Bluetooth, BT-LE, NFC, RFID, UWB, and the like. The radio interface may include other components, such as filters, converters (for example, digital-to-analog converters and the like), symbol demappers, signal shaping components, an Inverse Fast Fourier Transform (IFFT) module, and the like, to process symbols, such as OFDMA symbols, carried by a downlink or an UL.

[0072] For instance, transceivers 18 and 28 may be configured to modulate information on to a carrier waveform for transmission by the antennas 15 and 25 and demodulate information received via the antenna 15 and 25 for further processing by other elements ofapparatuses 10 and 20. In other example embodiments, transceivers 18 and 28 may be capable of transmitting and receiving signals or data directly. Additionally or alternatively, in some example embodiments, apparatus 10 may include an input and / or output device (I / O device). In certain example embodiments, apparatuses 10 and 20 may further include a user interface, such as a graphical user interface or touchscreen.

[0073] In certain example embodiments, memories 14 and 34 store software modules that provide functionality when executed by processors 12 and 22. The modules may include, for example, an operating system that provides operating system functionality for apparatuses 10 and 20. The memory may also store one or more functional modules, such as an application or program, to provide additional functionality for apparatuses 10 and 20. The components of apparatuses 10 and 20 may be implemented in hardware, or as any suitable combination of hardware and software. According to certain example embodiments, apparatuses 10 and 20 may optionally be configured to communicate each other (in any combination) via a wireless or wired communication links 70 according to any radio access technology, such as NR.

[0074] According to certain example embodiments, processors 12 and 22 and memories 14 and 24 may be included in or may form a part of processing circuitry or control circuitry. In addition, in some example embodiments, transceivers 18 and 28 may be included in or may form a part of transceiving circuitry.

[0075] For instance, in certain example embodiments, apparatus 10 may be controlled by memory 14 and processor 12 to receive symbols from a network element. Apparatus 10 may also be controlled by memory 14 and processor 12 to receive an array including a number of subcarriers, a number of the symbols, and a number of input channels. According to certain example embodiments, the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. Apparatus 10 may further be controlled by memory 14 and processor 12 to process the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, apparatus 10 may be controlled by memory 14 and processor 12 to jointly train the receiver weights and a transmitter constellation shapeassociated with a demodulation symbol.

[0076] In other example embodiments, apparatus 20 may be controlled by memory 24 and processor 22 to receive, from a user equipment, a channel quality indicator. Apparatus 20 may also be controlled by memory 24 and processor 22 to select, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. Apparatus 20 may further be controlled by memory 24 and processor 22 to adjust the modulation coding scheme and a demodulation reference signal density based on a block error rate.

[0077] In some example embodiments, an apparatus (e.g., apparatus 10 and / or apparatus 20) may include means for performing a method, a process, or any of the variants discussed herein. Examples of the means may include one or more processors, memory, controllers, transmitters, receivers, and / or computer program code for causing the performance of the operations.

[0078] Certain example embodiments may be directed to an apparatus that includes means for performing any of the methods described herein including, for example, means for receiving symbols from a network element. The apparatus may also include means for receiving an array including a number of subcarriers, a number of the symbols, and a number of input channels. According to certain example embodiments, the input channels may include a received signal, a raw channel estimate, or a demodulation reference signal density scalar value. The apparatus may further include means for processing the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols. In addition, the apparatus may include means for jointly training the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

[0079] Other example embodiments may be directed to an apparatus that includes means for performing any of the methods described herein including, for example, means for receiving, from a user equipment, a channel quality indicator. The apparatus may also include means for selecting, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density. The apparatus mayfurther include means for adjusting the modulation coding scheme and a demodulation reference signal density based on a block error rate.

[0080] Certain example embodiments described herein provide several technical improvements, enhancements, and / or advantages in respect to signaling for communicating with learned constellations and reduced pilots with ML receivers. For instance, according to certain example embodiments, it may be possible to outperform purely pilot-based systems due to the increased robustness of learned constellations. In other example embodiments, it may be possible to increase the throughput and spectral efficiency via the reduced DMRS overhead without having to implement completely pilotless transmissions and without having to rely on excessively large machine learning models for the receiver.

[0081] A computer program product may include one or more computer-executable components which, when the program is run, are configured to carry out some example embodiments. The one or more computer-executable components may be at least one software code or portions of it. Modifications and configurations required for implementing functionality of certain example embodiments may be performed as routine(s), which may be implemented as added or updated software routine(s). Software routine(s) may be downloaded into the apparatus.

[0082] As an example, software or a computer program code or portions of it may be in a source code form, object code form, or in some intermediate form, and it may be stored in some sort of carrier, distribution medium, or computer readable medium, which may be any entity or device capable of carrying the program. Such carriers may include a record medium, computer memory, read-only memory, photoelectrical and / or electrical carrier signal, telecommunications signal, and software distribution package, for example. Depending on the processing power needed, the computer program may be executed in a single electronic digital computer or it may be distributed amongst a number of computers. The computer readable medium or computer readable storage medium may be a non- transitory medium.

[0083] In other example embodiments, the functionality may be performed by hardware orcircuitry included in an apparatus (e.g., apparatus 10 or apparatus 20), for example through the use of an application specific integrated circuit (ASIC), a programmable gate array (PGA), a field programmable gate array (FPGA), or any other combination of hardware and software. In yet another example embodiment, the functionality may be implemented as a signal, a non-tangible means that can be carried by an electromagnetic signal downloaded from the Internet or other network.

[0084] According to certain example embodiments, an apparatus, such as a node, device, or a corresponding component, may be configured as circuitry, a computer or a microprocessor, such as single-chip computer element, or as a chipset, including at least a memory for providing storage capacity used for arithmetic operation and an operation processor for executing the arithmetic operation.

[0085] One having ordinary skill in the art will readily understand that the disclosure as discussed above may be practiced with procedures in a different order, and / or with hardware elements in configurations which are different than those which are disclosed. Therefore, although the disclosure has been described based upon these example embodiments, it would be apparent to those of skill in the art that certain modifications, variations, and alternative constructions would be apparent, while remaining within the spirit and scope of example embodiments. Although the above embodiments refer to 5G NR and LTE technology, the above embodiments may also apply to any other present or future 3GPP technology, such as LTE-advanced, and / or fourth generation (4G) technology.

[0086] Partial Glossary:

[0087] 3GPP 3rd Generation Partnership Project

[0088] 5G 5th Generation

[0089] 5GCN 5G Core Network

[0090] 5GS 5G System

[0091] BS Base Station

[0092] CQI Channel Quality Indicator

[0093] DL Downlink

[0094] DMRS Demodulation Reference Signal

[0095] eNB Enhanced Node B

[0096] E-UTRAN Evolved UTRAN

[0097] gNB 5G or Next Generation NodeB

[0098] LTE Long Term Evolution

[0099] MCS Modulation and Coding Scheme

[0100] ML Machine Learning

[0101] NR New Radio

[0102] OFDM Orthogonal Frequency-Division Multiplexing

[0103] OLLA Outer Loop Link Adaptation

[0104] QAM Quadrature Amplitude Modulation

[0105] Tx / Rx Transmitter / Receiver

[0106] UE User Equipment

[0107] UL Uplink

Claims

WE CLAIM:

1. A method comprising: receiving symbols from a network element; receiving an array comprising a number of subcarriers, a number of the symbols, and a number of input channels, wherein the input channels comprise a received signal, a raw channel estimate, or a demodulation reference signal density scalar value; processing the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols; and jointly training the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

2. The method according to claim 1 , further comprising: outputting a final log-likelihood ratio estimate based on the processed array.

3. The method according to claims 1 or 2, wherein the transmitter constellation shape is trained under a sparse demodulation reference symbol configuration.

4. The method according to any of claims 1-3, wherein the transmitter constellation shape is modified based on encoded bits and bit probabilities.

5. The method according to any of claims 1-4, further comprising: mapping a demodulation reference signal density to the transmitter constellation shape, wherein the demodulation reference signal density is associated with either a single constellation that supports varying demodulation reference signal densities, or a different predefined constellation for each demodulation reference signal density.

6. The method according to any of claims 1-5, further comprising: assigning a separate constellation for different modulation orders; and detecting the different modulation orders, wherein the detection is performed with separately trained receiver models, or with a single receiver model trained to support a plurality of constellation sizes.

7. A method, comprising: receiving, from a user equipment, a channel quality indicator; selecting, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density; and adjusting the modulation coding scheme and a demodulation reference signal density based on a block error rate.

8. The method according to claim 7, further comprising: selecting a modified constellation based on the adjustment of the demodulation reference density.

9. The method according to claims 7 or 8, further comprising: incorporating the demodulation reference signal density to an outer loop link adaptation algorithm via a modulation coding scheme table comprising modified constellations.

10. The method according to any of claims 7-9, further comprising: selecting the modulation coding scheme and the demodulation reference signal density based on a reported channel quality indicator, a reported error, a traffic type, or channel conditions.

11. An apparatus, comprising: at least one processor; and at least one memory comprising computer program code, wherein the at least one memory and the computer program code are configured to, with the at least one processor, cause the apparatus at least to receive symbols from a network element; receive an array comprising a number of subcarriers, a number of the symbols, and a number of input channels, wherein the input channels comprise a received signal, a raw channel estimate, or a demodulation reference signal density scalar value; process the array with one or more residual network blocks to obtain receiver weights corresponding to the received symbols; and jointly train the receiver weights and a transmitter constellation shape associated with a demodulation symbol.

12. The apparatus according to claim 11, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to: output a final log-likelihood ratio estimate based on the processed array.

13. The apparatus according to claims 11 or 12, wherein the transmitter constellation shape is trained under a sparse demodulation reference symbol configuration.

14. The apparatus according to any of claims 11-13, wherein the transmitter constellation shape is modified based on encoded bits and bit probabilities.

15. The apparatus according to any of claims 11-14, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to: map a demodulation reference signal density to the transmitter constellation shape,wherein the demodulation reference signal density is associated with either a single constellation that supports varying demodulation reference signal densities, or a different predefined constellation for each demodulation reference signal density.

16. The apparatus according to any of claims 11-15, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to: assign a separate constellation for different modulation orders; and detect the different modulation orders, wherein the detection is performed with separately trained receiver models, or with a single receiver model trained to support a plurality of constellation sizes.

17. An apparatus, comprising: at least one processor; and at least one memory comprising computer program code, wherein the at least one memory and the computer program code are configured to, with the at least one processor, cause the apparatus at least to receive, from a user equipment, a channel quality indicator; select, based on the channel quality indicator, an initial modulation coding scheme and an initial demodulation reference signal density; and adjust the modulation coding scheme and a demodulation reference signal density based on a block error rate.

18. The apparatus according to claim 17, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to: select a modified constellation based on the adjustment of the demodulation reference density.

19. The apparatus according to claims 17 or 18, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to: incorporate the demodulation reference signal density to an outer loop link adaptation algorithm via a modulation coding scheme table comprising modified constellations.

20. The apparatus according to any of claims 17-19, wherein the at least one memory and the computer program code are further configured to, with the at least one processor, cause the apparatus at least to: select the modulation coding scheme and the demodulation reference signal density based on a reported channel quality indicator, a reported error, a traffic type, or channel conditions.

Images