Method for determining at least one precoding matrix for implementing a payload data transmission from a transmitter to a receiver

Abstract

The invention relates to determining, by a receiver, a precoding for a transmission from a transmitter to the receiver. The method comprises, for at least one given number of spatial layers v, selecting (E1), from a beam dictionary (B), a number v + k of beams exhibiting, for example, the best performance with regard to a first performance criterion, where k is a strictly positive integer; and determining (E2) at least two distinct wideband precoding matrices each formed from v beams among the v + k selected beams.

Description

[0001] DESCRIPTION

[0002] TITLE: Method for determining at least one precoding matrix for implementing useful data transmission from a transmitter to a receiver

[0003] 1. Scope of the invention

[0004] The invention relates to the field of wireless communications in multiple input and multiple output or MIMO (“Multiple Input- Multiple Output” in English).

[0005] More specifically, the invention proposes a technique for optimizing the formation of beams obtained from an antenna array, in order to improve the transmission of information between a transmitter and a receiver, in both the uplink and downlink.

[0006] The invention has applications in any beamforming-based system, particularly in radio communication networks according to the 4G or 5G standards defined by 3GPP, WiFi communication networks according to the various IEEE 802.11 standards, etc.

[0007] For downlink communication, the transmitter can be a base station, for example an eNodeB (evolved Node B) for networks based on LTE or LTE Advanced technologies, or a Wi-Fi access point, etc. A receiver can be a device such as a smartphone, tablet, or connected object. For uplink communication, the transmitter can be a device, and the receiver a base station.

[0008] 2. Prior art

[0009] Beamforming, or precoding, is a signal processing technique used in antenna or sensor arrays for the directional transmission or reception of signals. In other words, thanks to antenna arrays, transmitters and / or receivers can focus the radiation of the emitted wave in a particular direction, thus achieving spatial selectivity.

[0010] Beamforming is achieved by combining the elements of a phase- and amplitude-controlled antenna array such that:

[0011] The signals combine constructively in particular directions, resulting in a strengthening of the useful power received, and

[0012] The signals combine destructively in other directions, resulting in a decrease in the received interference power. Therefore, to form beams at a transmitter, a complex coefficient, called a precoding coefficient, is applied to each element of the transmitter's antenna array. The set of these coefficients forms a precoding matrix.

[0013] Note that for the radiation pattern to be oriented in the desired direction, the precoding coefficients must be carefully chosen. The better the precoding, the better the transmission.

[0014] A fundamental problem in selecting this precoding is acquiring knowledge of the transmitting channel, or "Channel State Information" (CSI). In information theory, channel state knowledge refers to knowledge of the channel from transmitter to receiver and the statistics of interference in the receiver. Acquiring CSI allows us to characterize the channel's properties and determine the appropriate precoding matrix.

[0015] Currently, two techniques for acquiring transmit channel knowledge are proposed for MIMO systems in the 4G, 5G, and IEEE 802.11x (IEEE 802.11n, 802.11ac, 802.11ax) standards: a technique denoted CSI-D based on the use of precoding dictionaries or directories ("codebooks") and a technique based on channel reciprocity ("channel reciprocity"). The channel knowledge acquisition techniques, and the associated reference signals, are described in more detail in the 3GPP specifications TS36.213 and TS36.211 for 4G and TS38.211 and TS38.214 for 5G.

[0016] The CSI-D technique relies on finite precoding dictionaries defined by a standard (such as the 4G or 5G standard) and the use of a limited return channel between the receiver and transmitter. Through this channel, the receiver indicates to the transmitter a preferred precoding choice from among the dictionary elements. CSI-D is particularly advantageous in scenarios where reciprocity between the transmit and receive channels is not applicable (for example, in Frequency Division Duplex (FDD) systems), or where the receiver is located at the cell edge, where interference is significant and the receiver's transmission conditions are potentially degraded (for example, due to limited transmission power).

[0017] More specifically, according to this CSI-D technique, the transmitter emits a reference signal, also called a pilot signal. Such a reference signal is typically denoted CSI-RS in 4G and 5G standards, for "Channel State Information - Reference Signal". Upon receiving the reference signal, the receiver estimates, on the one hand, the transmission channel between the transmitter and the receiver (i.e., in the transmitter-to-receiver direction), and on the other hand, an interference covariance matrix, representative of the spatial structure of the interference between the receiving antennas. From the estimation of the transmission channel between the transmitter and the receiver and the spatial characteristics of the interference, the receiver determines the precoding matrix to be used by the transmitter from the finite precoding dictionary for the transmission of useful data to the receiver.To this end, the receiver can use a physical layer abstraction that allows it to simulate theoretical behavior with a given precoding scheme, in order to compare the performance between different precoding schemes and thus predict the transmission format that maximizes a certain criterion, typically the data rate for the considered frequency sub-band. By transmission format, we mean a precoding matrix and a corresponding coding and modulation scheme.

[0018] The receiver then sends this precoding choice back to the transmitter via the limited return channel, for example in the form of a Precoding Matrix Indicator (PMI). The precoding choice can optionally be accompanied by a Channel Quality Indicator (CQI) and / or a Rank Indicator (RI) indicating the number of spatial layers.

[0019] In this context, the 3GPP standard defines several types of dictionaries: Type I dictionaries and Type IL dictionaries

[0020] Type I dictionaries ensure good connectivity while being energy-efficient, but at the cost of lower channel estimation accuracy. Type II dictionaries, on the other hand, allow for more precise channel estimation, which is particularly useful when the receiver is not in favorable radio conditions (for example, at the edge of a cell), but at the cost of increased computational overhead at the receiver and more signaling traffic back to the transmitter (which always comes at the expense of payload transmission). A mobile terminal can therefore switch between using Type I or Type II dictionaries, depending on its connectivity over time.

[0021] In this context, 3GPP version 19 (release 19, or abbreviated as rel-19) offers a refinement of the Type I dictionary. This refinement aims, in particular, to handle scenarios with more than 32 antenna ports on the base station side, for example, 48, 64, or 128 antenna ports. Type I dictionaries are divided into two modes, known as Mode A and Mode B. Mode A extends the legacy mechanisms of the Rel-15 Type I codebook to antenna port counts higher than 32. Mode B reuses the legacy mechanisms of the Rel-16 eType II codebook to independently determine the precoding vector for each spatial layer. The decision and description of these two modes (or schemes) are detailed in the "Final minutes report" of the "chairman's notes" of meeting RAN#116 Bis, available at the following URL:

[0022]

[0023] 116b vl00.zip.

[0024] As part of this refinement, a precoding matrix is ​​formed by multiplying two components:

[0025] a broadband precoding matrix, denoted W1, common to all sub-bands, which defines a set of beams to be used at the transmitter; and sub-band precoding matrices, each specific to a given sub-band n and denoted W2 n , which define how the beams defined in the matrix

[0026]

[0027] are combined for this given sub-band n.

[0028] A precoding matrix for a given sub-band n can then be written in the form W n = W1W2. By abuse of language, we sometimes write W = W1W2 to define the resulting precoding.

[0029] The precoding selection strategy according to Release-19 type-l mode B then comprises two steps: a first step in which the broadband precoding matrix

[0030]

[0031] common to all sub-bands is selected, and a second step in which the coefficients for each sub-band (i.e. the W2 matrices specific to each sub-band) are selected.

[0032] In the first step, as defined in the refinement according to Release-19 type-l mode B, the matrix is ​​formed by selecting, from a dictionary of B beams defined by the standard, the v best beams in terms of power received by the receiver, where v is the number of spatial layers associated with the transmission. More specifically, this selection is performed by sequentially determining v beams, each beam being selected independently from among the beams not yet selected, so as to maximize the received power while taking into account both polarizations.

[0033] This method of determining the broadband precoding matrix

[0034]

[0035] However, this is not without its drawbacks. Indeed, this independent and sequential selection of the beams constituting the broadband precoding matrix

[0036]

[0037] does not take into account the risks of interference between spatial layers. However, in some cases, the combination of the best beams (in terms of received power) induces strong interference likely to significantly degrade the quality of reception (SINR (“Signal-to-interference-plus-noise ratio”) and throughput), leading such a combination to ultimately be suboptimal in terms of throughput obtained.

[0038] Therefore, there is a need for a solution that can determine precoding without the aforementioned drawbacks, while remaining sufficiently accurate in terms of channel estimation, and sufficiently simple to be implemented even on mobile user equipment with limited computing power. 3. Description of the invention

[0039] The invention improves the situation. To this end, the invention proposes a method for determining, by a receiver, at least one precoding matrix, for implementing a transmission from a transmitter to said receiver through a transmission channel, said method comprising, for at least a given number of spatial layers v:

[0040] the selection, within a dictionary of sheaves (B), of a number v + k of sheaves, with k a strictly positive integer; and

[0041] the determination of at least two broadband precoding matrices (

[0042]

[0043] W1 power W1 inter ) distinct, each formed from v beams among said v + k selected beams.

[0044] For example, the selected v + k beams are those that exhibit the best performance within the beam dictionary with respect to a first performance criterion.

[0045] In a particular mode of transmission, the determination process may further include:

[0046] the transmission, to said sender, of at least one piece of information (PMI) identifying the broadband precoding matrix, among said determined broadband precoding matrices, defining a precoding (W) exhibiting better performance with regard to a second performance criterion, compared with at least one precoding defined by another of said determined broadband precoding matrices.

[0047] Thus, thanks to this novel process, the selected broadband precoding matrix, which serves as the basis for constructing the precoding implemented at the transmitter, takes into account both a first performance criterion (for example, receive power) related to the beams that compose it, and other data that can influence a second performance criterion (for example, the data rate of the selected precoding, or its SINR), such as beam interaction, i.e., interference between spatial layers. This makes it possible to improve the second performance criterion (for example, the receive data rate or the SINR associated with this precoding) since the "naive" selection of the broadband precoding matrix (i.e., the one defined by the v best beams (in the sense of the first performance criterion) selected sequentially and independently of each other) does not maximize this second performance criterion.

[0048] Furthermore, since the beams constituting this low-interference broadband precoding matrix are selected from a subset B v+K of reduced size (v + K, K > 0) within the dictionary, the computational complexity of the process is considerably lower compared to an exhaustive determination of all possible combinations of sheaves from dictionary B in order to maximize the first performance criterion. This subset B v+K Since the reduced size consists of the "best" beams (according to the first performance criterion), this nevertheless ensures that each beam exhibits good performance (in terms of the first performance criterion, for example, satisfactory receive power or a satisfactory SINR). Thus, the matrix(s) constructed from this subset of best beams B v+Kare all based on beams exhibiting "good" performance with respect to this first performance criterion (e.g., receive power). The first performance criterion (related to beam selection in the dictionary) influences the second performance criterion (related to the transmitted precoding matrix), ensuring that the transmitted precoding matrix performs well while being more economical to determine than with prior art techniques.

[0049] This compromise between maximizing the second performance criterion and reducing the number of combinations to be tested allows the process to be both more economical and precise than prior art techniques.

[0050] According to one particular aspect, the first performance criterion, on the basis of which the v + k beams are selected in the beam dictionary, is a receiving power at the receiver level.

[0051] According to one particular aspect, the second performance criterion, on the basis of which information (PMI) representative of a precoding matrix is ​​transmitted, is a receive rate.

[0052] According to a particular aspect, said at least two broadband precoding matrices include at least one broadband precoding matrix {W^ ower ) formed from the v beams exhibiting the best performance with regard to said first performance criterion.

[0053] According to one particular aspect, the determination of a plurality of precoding matrices is implemented for all distinct broadband precoding matrices formed from the selection of v beams from said v + k beams, and wherein the broadband precoding matrix whose identifier is transmitted to the transmitter is the one whose resulting precoding has a better second performance criterion than that of any other precoding formed by any of the other broadband precoding matrices.

[0054] In such a case, called "exhaustive," all combinations of broadband precoding matrices are tested. This allows the best (e.g., in terms of throughput) of these broadband precoding matrices to be determined with certainty, while maintaining a low computational cost since only the v + k best vectors from the dictionary are considered. According to a particular aspect, the determination of a plurality of precoding matrices is implemented for two distinct broadband precoding matrices formed from the selection of v bundles from said v + k bundles, comprising:

[0055] a first broadband precoding matrix corresponding to said broadband precoding matrix (W1 power ) formed from the v beams associated with better reception power at said receiver;

[0056] a second broadband precoding matrix (VK) lter), formed from the v beams minimizing interference between spatial layers.

[0057] In this case, the complexity of selecting a broadband precoding matrix is ​​greatly reduced, because only two of these matrices are selected, including the "classical" broadband precoding matrix, the determination of which is computationally economical.

[0058] According to one particular aspect, the second broadband precoding matrix (W nter ) is determined to be the broadband precoding matrix (

[0059]

[0060] W^), among all the distinct broadband precoding matrices formed from the selection of v beams from said v + k beams, minimizing the interference defined by the following formula:

[0061] p

[0062] h with p = 1.2

[0063]

[0064] where | | p represents the norm p, bn (i) represents the i-th coefficient of the beam b n taken from said v + k bundles, and R 11 and R 22 are the diagonal blocks of the interference covariance matrix R.

[0065] According to one particular aspect, the transmission to said sender includes the joint transmission of an identifier associated with the first broadband precoding matrix (W^ ower ) and an identifier associated with the second broadband precoding matrix (W1 inter ).

[0066] According to one particular aspect, the said joint transmission is conditional upon prior verification of a configuration data previously received from said sender.

[0067] According to one particular aspect, the said number k is a data received from the said sender, prior to the said selection.

[0068] According to one particular aspect, said number k is of the same order of magnitude as a maximum number of spatial layers (y max ).

[0069] By "of the same order of magnitude," we mean that the ratio k / v does not exceed 10, preferably not 5, and even more preferably 2. In a preferred embodiment, k < v, that is to say, the number of beams considered is at most doubled compared to a classical method in which only the v best beams are retained to determine W. This makes it possible to greatly reduce the complexity (in the computational sense) of the calculation of the broadband precoding matrix returned to the transmitter.

[0070] The invention also relates to a communication method for implementing a transmission from a transmitter to a receiver through a transmission channel, said method being implemented by the transmitter and comprising, for at least a given number of spatial layers v:

[0071] a reception, from the receiver, of at least a second identifier associated with a second broadband precoding matrix (W1 inter ) minimizing beam interference, the second broadband precoding matrix being determined on the basis of a set of v + k beams included in a beam dictionary (B), with k a strictly positive integer; and an implementation of precoding on the basis of a precoding matrix function of at least one identifier associated with the second broadband precoding matrix.

[0072] For example, the beams of said set of v + k beams are those which have the best performance with regard to a first performance criterion within the beam dictionary (B).

[0073] By receiving the identifier from a broadband precoding matrix of the type mentioned above, the transmitter can implement precoding for transmission to the receiver with reduced beam interference. This allows, in many situations, for improved receive throughput and / or SINR, and therefore improved communication quality.

[0074] According to a particular aspect of this transmitter-side communication process, this process includes, in conjunction with said reception of at least a second identifier associated with the second matrix, a reception, from the same receiver, of a first identifier associated with a first broadband precoding matrix (W1 power ) determined on the basis of the v bundles of dictionary (B) which maximize the first performance criterion.

[0075] By jointly receiving the identifiers of the first and second wide precoding matrices of the claimed type, the transmitter also has information on the best directions (i.e., the beams) in terms of the first performance criterion (e.g., in terms of received power for a beam), as well as the best directions (i.e., the best beam combination) minimizing interference among the best beams in terms of the first performance criterion. This information combined allows the transmitter (like a base station) to better understand the environment, which is particularly advantageous for MU-M1 MO transmission.

[0076] According to one particular aspect, the process further includes, prior to the receipt of at least a second identifier, the transmission of a control message comprising configuration data requiring a joint transmission of said identifiers of said first and second broadband precoding matrices.

[0077] The invention further relates to a receiver configured to determine at least one precoding matrix for implementing a transmission from a transmitter to said receiver through a transmission channel, said receiver comprising at least one processor configured for at least a given number of spatial layers v:

[0078] select, within a dictionary of sheaves (B), with k a strictly positive integer; and

[0079] determine at least two broadband precoding matrices (W^ ower ; JV™ ter) distinct, each formed from v beams among said v + k selected beams.

[0080] For example, the selected v + k beams are those that exhibit the best performance with regard to a first performance criterion within the beam dictionary (B).

[0081] In a particular embodiment the processor is further configured to transmit, to said transmitter, at least one piece of information (PMI) identifying the broadband precoding matrix, among said determined broadband precoding matrices, defining a precoding (W) exhibiting better performance with regard to a second performance criterion, compared with at least one precoding defined by another of said determined broadband precoding matrices.

[0082] The invention further relates to a transmitter configured for implementing a transmission from said transmitter to a receiver through a transmission channel, the transmitter comprising at least one processor configured for at least a given number of spatial layers v:

[0083] receive, from the receiver, at least a second identifier associated with a second broadband precoding matrix W1interference minimizing beam interference, the second broadband precoding matrix being determined on the basis of a set of v + k beams included in a beam dictionary (B), with k a strictly positive integer; and

[0084] implement precoding based on a precoding matrix that is a function of at least one identifier associated with the second broadband precoding matrix.

[0085] For example, the beams of said set of v + k beams are those which have the best performance with regard to a first performance criterion within the beam dictionary (B).

[0086] The invention further relates to a program and a computer program product comprising instructions for implementing one of the aforementioned processes when this program is executed by a processor. 4. List of Figures

[0087] Other features and advantages of the invention will become clearer upon reading the following description of illustrative and non-limiting embodiments and the accompanying drawings, among which:

[0088] [Fig. 1] schematically represents a communication system, in a particular embodiment of the proposed technique;

[0089] [Fig. 2] schematically represents a method for determining at least one precoding matrix, in a particular embodiment of the proposed technique;

[0090] [Fig. 3] describes a simplified architecture of a receiving equipment for the implementation of the proposed technique, in a particular embodiment;

[0091] [Fig. 4] describes a simplified architecture of a receiving equipment for the implementation of the proposed technique, in a particular embodiment.

[0092] 5. Detailed description

[0093] 5.1. Object of the invention

[0094] The invention relates to an improvement in the techniques for determining a broadband precoding matrix at the level of a receiving device (hereafter referred to simply as "receiver"). By way of illustration, and without limitation, such an improvement can, for example, be implemented as an alternative to the technique for determining a broadband precoding matrix as currently proposed in version 19 of the 3GPP standard, in connection with the proposed refinement of the Type I Mode B dictionary. As already described in relation to the prior art, such a broadband precoding matrix determined by the receiver is intended to be used by a transmitting device (hereafter referred to simply as "transmitter") for the transmission of useful data from the transmitter to the receiver.

[0095] As introduced in relation to prior art, by broadband precoding matrix, usually denoted W lt A precoding matrix is ​​understood to be a precoding matrix common to several sub-bands or to all sub-bands, which defines a set of beams to be used at the transmitter. A precoding matrix relative to a sub-band is understood to be a precoding matrix specific to a given sub-band n, denoted

[0096]

[0097] defining how the beams defined in the matrix

[0098]

[0099] are combined for this given sub-band.

[0100] 5.2. Implementation context In this context, in relation to Figure 1, we present an example of a transmission system in which the present technique can be implemented and used, in a particular embodiment.

[0101] Such a transmission system 1 comprises a transmitter 2 and a receiver 3, communicating via a channel 4. The transmitter 2 comprises a number N T 5 transmitting antennas (N T > 1). The receiver 3 comprises a number N R 6 receiving antennas (N R > 1).

[0102] In one aspect, the invention relates to the determination by the receiver 3 of a precoding to be applied by the transmitter 2 for the transmission of useful data to the receiver 3, so that this transmission takes place under the most optimal conditions possible (minimizing information loss, maximizing data rate, etc.). "Useful data" here refers to the data that constitutes the message that a user of the transmitter 2 wishes to communicate to a user of the receiver 3, as opposed, for example, to "signaling" or "service" data necessary only for the technical implementation of the communication. Such a precoding is determined from a set of predefined precodings known to both the transmitter 2 and the receiver 3, the set being of finite size and defined, for example, by a standard. This precoding, once determined at the receiver 3, is sent back to the transmitter 2 via a limited return channel.Transmitter 2 can then implement a transmission by applying this precoding. In this context, the information sent from receiver 3 to transmitter 2 typically includes a precoding matrix indicator (PMI). It may also include indicators such as a rank indicator (RI) and / or a channel quality indicator (CQI).

[0103] As illustrated in Figure 1, the transmitter and receiver communicate through a channel H, presenting for each sub-band n of an N3 number of frequency sub-bands a channel matrix H n , i.e. H = {H n} r ^ 3 =1 The number of spatial layers v is selected by the receptor according to the channel H = [H n}^ 3 =1for example, to maximize throughput for a target block error rate (generally 10%) and cannot exceed v max = mln(N T , N R ),

[0104] Each sub-band n is also associated with a covariance matrix of noise plus interference (denoted K n for a given sub-band n), representative of the spatial structure of the interference in reception for that sub-band. The covariance of the channel before bleaching is called the matrix R c = H nH n , where N3 is the number of sub-bands of the H channel n , and covariance of the

[0105]

[0106] N3

[0107] channel after bleaching matrix R = HnK~ H n - The receptor is able to estimate the channel

[0108]

[0109] H, by sub-band, using reference signals (CSI-RS) transmitted on the different sub-bands by the transmitter. Here, the structure of the interference refers to the correlation of the interference on the different receiving antennas for a subcarrier (given frequency) of an OFDM symbol, for example. The interference will be described as "structured" when the interference covariance matrix K n deviates from an identity matrix (up to a multiplicative factor), that is, when the correlation between the receiving antennas is strong: the non-diagonal elements of K n are not negligible compared to the diagonal elements, i.e., the matrix K n moves away from a diagonal matrix.

[0110] The receiver here having several receiving antennas, the noise plus interference covariance matrix is ​​representative of the spatial structure of the interference between the receiving antennas.

[0111] Estimating this noise-plus-interference covariance matrix for a sub-band relies on a configuration of CSI interference measurements (or CSI-IM). These CSI-IMs indicate resource element (RE) positions where nothing is being transmitted by the transmitter, thus providing a transmitter-free interference measurement window. In other words, the time-frequency positions (i.e., resource elements) where the covariance is to be estimated correspond to zero-power CSI-RSs, meaning resource elements that are not being used for transmission.

[0112] The estimation can be performed by correlating representative received signals of the interference on the different receiving antennas 6, in the absence of CSI-RS reference signals. The transmitter 2, for example, can configure time-frequency resources prohibited for transmission (a technique known as "Zero-Power CSI-RS" in the 3GPP ZP-CSI-RS TS38.211 standard), which allows the receiver 3 to more easily measure the interference on these resources for each receiving antenna 6. Alternatively, the noise-interference covariance matrix can be estimated by subtracting the useful data or reference signals—that is, those not considered part of the interference from the perspective of the receiver measuring it—from the signals received on each antenna to obtain signals representative of the interference on the different receiving antennas 6.

[0113] A broadband covariance matrix R of the channel after noise plus interference whitening is then defined, by averaging the matrices K o ,..., K N 3-1 sub-bands 0 to N3—1, according to the following formula, where

[0114]

[0115] denotes the adjoint matrix (i.e., the conjugate matrix of the transpose matrix) of H n

[0116] N3-I

[0117] n _ J_ 'v _ r^n R 12

[0118] N3ZU " NAN N >21 «22I

[0119]

[0120] n=0

[0121] 5.3. General principle of the invention, particular embodiments The general context of implementation of the invention having now been established, its general principle and various particular embodiments are presented below.

[0122] More particularly, the general principle of the invention is based on an extensive selection, within a dictionary of beams, of the number of beams usable to form the broadband precoding matrix 14^. In other words, such a selection is extended strictly beyond the classical selection (for example, as defined in the refinement according to version 19 type-l mode B of the 3GPP standard) of the v best beams in terms of power received by the receiver (where v is the number of spatial layers) to form the broadband precoding matrix.As detailed below, such an extended selection makes it possible in particular to explore a plurality of beam combinations for the formation of a broadband precoding matrix, and thus to determine a broadband precoding matrix which potentially has better characteristics, in terms of minimizing interference between spatial layers and therefore throughput, than the classically used broadband precoding matrix.

[0123] In other words, the general principle of the invention, for determining a broadband precoding matrix, relies on a combination of selecting the best beams (with respect to a first performance criterion) and an exhaustive search for a good beam combination among a large number of candidate broadband precoding matrices. The invention proposes an assumption according to which the best broadband precoding matrix (with respect to a second performance criterion, for example, in the sense of maximizing the effective data rate / SI NR for a target block error rate or minimizing the average interference between beams evaluated according to an approximation) is formed by v beams taken from among the v + k best beams, where k is a strictly positive integer.

[0124] This first performance criterion could, for example, be the reception power without taking into account the spatial structure of the interference based on the channel covariance before bleaching R c , i.e., find the v + k beams v t that maximize i=i [v 7

[0125]

[0126] [ y l ] or alternatively, the power in reception taking into account the interference structure based on the broadband covariance of the channel after noise whitening plus interference R, which consists of finding the v + k beams v t which maximize Sï=i k [v 7 v^]R [y 1 ])- *- e calculation with the matrix R c is less precise, but simpler to calculate. Note that this first performance criterion does not take into account beam interference.

[0127] Based on this hypothesis, the invention proposes, firstly, to select more beams than necessary, i.e., strictly more beams than the v spatial layers, contrary to what is proposed in the prior art. In the following description, by way of illustration and not limitation, we consider the case where the first performance criterion (considered during beam selection) is the maximization of received power taking interference into account, and the second performance criterion (considered for evaluating the performance of a broadband precoding matrix) is the data rate. The v + k beams thus selected are then those with the highest total received power. By selecting the v + k beams with the highest total received power, we mean selecting v + k beams whose respective total received power is greater than that of unselected beams in the beam dictionary.Then, among these selected beams maximizing the received power, several combinations of beams, i.e. several broadband precoding matrices, are tested, so as to select the best of these broadband precoding matrices, that is to say the one which allows for example to maximize the throughput when transmitting useful data from the transmitter to the receiver.

[0128] We now present in more detail the determination of a precoding in a particular embodiment of the invention.

[0129] We consider the case where the issuer applies the CSI-D method described above, i.e. applies a precoding based on one of the precoding matrices from a dictionary of precodings defined by the standard.

[0130] Furthermore, we are situated within a framework where, as described in relation to prior art, a precoding matrix W n for a given sub-band n, where N3 is the number of sub-bands in the H channeln is defined as the following product:

[0131] V

[0132]

[0133] ne [0, N3- 1], w® = w ± .

[0134] It is understood here that the matrix W is common to all sub-bands, while each matrix est specific to a given sub-band. For simplicity, we can write W = W ± . W2- This two-step precoding construction method is used for example in the 3GPP Rel-19 type-1 mode B standard.

[0135] For the rest of the request, we call

[0136] - the list W = (W n ) ne[Q , N3-1} a precoding (also noted W = (IV 1; (W^)„ e[0 V3-1] )), the IVÎJ matrix une precoding matrix for sub-band n, and

[0137] the 14^! matrix a broadband precoding matrix.

[0138] The receiver's precoding process involves two steps: determining a suitable broadband precoding matrix and determining subband precoding matrices. The goal of this precoding determination is to maximize the receiver throughput while minimizing the computational cost.

[0139] The matrix

[0140]

[0141] is defined from a game v = (v m ) n=1 of v beams, according to the formula: Vi V v 0 W1J v 2X1 ^N t N2Xl g ^2N1N2X2v WI((V„X =1 ) =

[0142]

[0143] QN ± N2X1 OjV^Xl V 1 V v We define v power like the function which, for a given integer (for example the number of spatial layers v), returns the list v power(B) of the l beams of the DFT B dictionary exhibiting the highest received power, that is:

[0144] i

[0145] Vpowerd') *^rg TÏIOX y [a, a, ]7Î [_ 1,

[0146]

[0147] (a^Li 6 ® AJ Ld i J or, under the R approximation 21 = R^2= ^N1N2XN1N2I meaning that the bleached channels of the two polarizations are not correlated,

[0148] i

[0149] power(. ( m ) m=i arg max y a, (T^ii + ^22) ^1

[0150]

[0151] Alternatively, it is possible to disregard the interference, in which case the matrix R is replaced by the matrix R c (i.e., the covariance before interference whitening) in the formula above.

[0152] 5.3.1 Selection of the broadband precoding matrix

[0153] Reference is now made to Figure 2 to illustrate, in a particular embodiment of the present technique, the method of determining at least one precoding matrix for the implementation of a useful data transmission from a transmitter to a receiver.

[0154] The inventors observed that, for certain channels

[0155]

[0156] The residual interference after selection according to the best beam criterion is too high, resulting in a loss of throughput. In other words, the matrix w ower

[0157]

[0158] = defined from the v beams offering the greatest received power, presents a resulting interference that impairs transmission performance.

[0159] To achieve a better compromise between received energy and interference between spatial layers, the invention then proposes in a step El to determine b =

[0160]

[0161] = v pO wer(y + the extended orthogonal basis from the DFT vector dictionary B, with k, as described previously, a strictly positive integer. This basis forms a set of "candidate" sheaves among which one or more subsets v 1 ,..., v 1 ... of v beams are selected so as to determine matrices = W (v 1 ) one of which has a better throughput than the others. These broadband precoding matrices are hereafter referred to as candidate matrices W^\ and their resulting precoding is called candidate precoding. To do this, several candidate matrices are determined in a step E2, according to strategies that will be detailed below.

[0162] Flow rate D Lassociated with a candidate matrix

[0163]

[0164] can be determined by abstracting from the physical layer according to the following formula, the details of which are explained below:

[0165] Dt = D (MCS, \siNR n (w^, W^, H n , tfjp

[0166]

[0167] The flow rates are thus calculated for all candidate matrices, which allows the matrix to be determined

[0168]

[0169] = IV^ 0 " 1 maximizing the throughput among the matrices considered, i.e.

[0170]

[0171] i Q = max Di for a given number of spatial layers v.

[0172] i

[0173] The candidate matrix

[0174]

[0175] presenting the best throughput is returned to the sender in an E3 step, in the form of a PMI.

[0176] To ensure reduced computational complexity, the parameter k defining the number v + k of determined beams is preferably at most of the same order of magnitude as the number of spatial layers v. Alternatively, particularly when small numbers v of spatial layers are tested, the parameter k is preferably at most of the same order of magnitude as the maximum number of spatial layers, that is, of the same order of magnitude as v max = min(N T , N R ).

[0177] By the same order of magnitude, we mean the ratio k / v maxdoes not exceed 10, preferably not 5 and even more preferably 2. In a preferred embodiment, k < v, that is to say that the number of beams considered is at most doubled compared to a classical method in which only the v best beams are retained to determine W.

[0178] This parameter k can be fixed, or change dynamically. It can be a function of the number of spatial layers considered, or conversely be agnostic to the number of spatial layers tested.

[0179] 5.3.2. Selection of sub-band precoding matrices and data rate calculation

[0180] The matrix

[0181]

[0182] P ourEach subband n, with n = 0, ..., N3—1, is selected from an exhaustive search of all possible subband matrices. This exhaustive search uses a physical layer abstraction to predict the data rate conditioned on a coding and modulation scheme (MCS), given a precoding choice W = W ± W2 and a signal-to-noise ratio plus associated interference SINR n (W, (W2) n , H n , Gi)> avec n e

[0183]

[0184] [1 — ~ 1] °ù K n the noise plus interference covariance matrix for the aforementioned sub-band n and H nThe channel matrix for sub-band n. Using a physical layer abstraction technique, such as that presented in the documents "Link performance models for system level simulations of broadband radio access systems" (K. Brueninghaus et al., IEEE 16th Int. Symposium on Personal, Indoor and Mobile Radio Communications, 2005 (PIMRC 2005), vol. 4, 2005, pp. 2306-2311) or "Realistic Performance of LTE: In a Macro-Cell Environment" (B. Landre et al., Proc. IEEE VTCS-2012, Japan, Yokohama, May 2012), a modulation and coding scheme can be determined based on this effective SINR and a target error rate using Gaussian quality tables. A method for calculating the effective SINR is detailed in this article, based on the SINRs. n mentioned above.

[0185] These Gaussian quality tables allow for the empirical correlation of the effective SINR, a target BLER error rate, and a modulation and coding scheme. The modulation and coding scheme, denoted MCS, allows for the calculation of a data rate. This data rate is determined from the modulation (number of bits per symbol) and the coding rate (ratio of useful bits). In one embodiment, a target BLER error rate of 10% (or more generally less than 20%) is set, which allows the modulation and coding scheme to be deduced, knowing the effective SINR. This then allows for obtaining a data rate for a precoding W = (W lr (Wf^n) - Formulated in mathematical terms and in this specific case:

[0186] = argmax I Q (siNR n (w^

[0187]

[0188] where IQ is the mutual information for the highest order modulation (256 QAM).

[0189] The flow rate is then calculated as follows:

[0190] D (MCS, [siNR n (w lt H n , K^ 3 ° )

[0191]

[0192] This formula is based on the SINR database. n per sub-band, a given modulation and coding scheme (MCS) can be used to deduce a channel quality indicator (CQI), a number of spatial layers v (or RI), and the pre-coding matrices (PMI). The format of

[0193]

[0194] L Jn=0

[0195] The transmission process will consist of choosing the CQI, PMI, RI that maximizes the predicted throughput for a target BLER of 10%. These matrices ] IV,} can thus be presented as a plurality of indicators.

[0196]

[0197] <■ Jn=0

[0198] PMI, like W.

[0199] We now describe below different strategies for determining the tape precoding matrix, in various particular embodiments of the present technique.

[0200] 5.3.3. Strategy for determining the tape precoding matrix by exhaustive search. In a first example implementation, all candidate matrices H > ) I=1 ..„- N = C "are considered, that is to say all combinations of v beams of the extended basis b =

[0201]

[0202] v p ower(y + are tested. The throughput of each of these candidate matrices

[0203]

[0204] is calculated, and the best of these matrices (in terms of throughput) is retained as the broadband precoding matrix.

[0205] This first selection strategy, called exhaustive search, ensures that, among the candidate matrices, the one with the best throughput will be returned to the sender. Since the number of combinations to be tested is C^ +k The computational complexity remains reasonable, which allows for the identification of a broadband precoding matrix that is better than the classically chosen matrix uz p0Vl ' er ( s j ce This matrix exists. Since this exhaustive search strategy explores the combinations of the best v + k beams in terms of received power, it is plausible that the best broadband precoding matrix retained is indeed the best broadband precoding matrix among all combinations in dictionary B, while being considerably faster to implement than an exhaustive search directly within dictionary B.

[0206] 5.3.4. Reduced-Complexity Tape Precoding Matrix Determination Strategy To reduce the complexity associated with the number of combinations to consider for a large k, the inventors identified a more economical strategy than the exhaustive search strategy described above. This second strategy, called interference minimization, consists of comparing j power to a candidate matrix lV^ nter q U j minimizes interference.

[0207] More specifically, to determine the matrix W" lter minimizing interference, we consider the matrix

[0208]

[0209] described above for i = 1,..., C^ +K In a first approximation, the average interference I k between spatial layers can be written

[0210] p

[0211] h with p = 1 or 2

[0212]

[0213] L ^m,n

[0214] Or} is a matrix concatenating a combination of v sheaves from the extended dictionary b, and | | p represents the norm p, corresponding to the Euclidean norm for p = 2 and the sum of the absolute values ​​of the coefficients for p = 1.

[0215] Under the

[0216]

[0217] 'hypothesis R' 21 = R^2= N xN N , the magnitude I k simplifies into

[0218] P

[0219] k = £m*n [< ed(0H (ff 11 +ff22X ed(0 l with p = 1.2

[0220]

[0221] L J m,n

[0222] or X" d<a = [6®... 6®].

[0223] The interference between the spatial layer m and n then simplifies to [x[ ed(l)H (ffll+ff22Xi ed(l) ] = av eC HW and (n, m) £ {l,---,v} 2 .

[0224]

[0225] And we then obtain the equation:

[0226] p

[0227] k with p = 1.2

[0228]

[0229] The interference minimization strategy then consists of determining the matrix W" lter minimizing interference, i.e. the combination j n = arqmin

[0230] ie[i. c +K ]

[0231] Within the framework of this strategy for determining the reduced-complexity tape precoding matrix, several specific embodiments can also be implemented for the transmission of precoding information from the receiver to the transmitter.

[0232] In the first case, the receiver determines the matrix W ter and the matrix w^ ower The receiver calculates their respective flow rates according to the abstraction of the physical layer described above, and compares the two flow rates. The receiver then returns to the transmitter the matrix that maximizes the flow rate.

[0233] In a second case, which is more economical in terms of computation time, the receiver determines the matrix W" lter and the matrix w ower e t returns the two matrices W" lter and w ower to the sender without comparing their data rates. This return can take the form of a transmission of two PMIs, one for each of these matrices. One of the two matrices w ter and

[0234]

[0235] V^ ower defining a precoding with a higher throughput than the precoding defined by the other matrix, the retrieval of the two matrices implies that one of the two matrices defines a better precoding than the other. The joint retrieval of the two matrices w ower e t yinter p erme td eCommunicating different information to the transmitter (i.e., the base station) involves, on the one hand, the best directions in terms of received power and, on the other hand, the best directions taking interference into account. These two pieces of information together allow the transmitter (i.e., the base station) to better understand the environment, which is advantageous for MU-MIMO transmission, for example.

[0236] The choice of the nature of the return (i.e., the two matrices Wf) ter and w^ ower , so The best of the two options in terms of throughput can be predefined or imposed by the sender. If the sender imposes this choice, the sender can transmit a control message to the receiver containing information indicating which of the two CSI return options is expected, based on an RRC configuration sent by the sender.

[0237] This interference minimization method, since it is based on candidate matrices resulting from combinations of v beams from among the v + k best beams, allows the selection of a matrix Wf ter exhibiting low interference and high received power. These two conditions are necessary for the W matrix ter either a relevant candidate (with an interference minimization strategy that would be implemented on the entire dictionary B, i.e. without pre-selection of the best v + k beams, the interference would certainly be very low, but so would the received power).

[0238] 5.3.5. Other strategies

[0239] Other strategies for selecting the best broadband precoding matrix (in the sense of throughput) from among candidate broadband precoding matrices (i.e. from combinations of v beams from among the v + k best beams) are conceivable.

[0240] For example, a so-called random strategy consists of randomly selecting one or more candidate matrices and comparing them to w ower e possibly at Wf ter , in the manner of a Monte Carlo method. Other strategies are conceivable to explore the possible combinations of v beams among the v + k best beams whose resulting matrix has a better throughput than the "naive" solution j P° wer

[0241] Strategies for selecting the best broadband precoding matrix for a given number of spatial layers have been described above. This selection can be performed for a plurality of spatial layer counts, or even all spatial layers between 2 and v maxThis allows for optimization of the spatial layer number criterion. In such a case, the result of the precoding dictionary selection process can be returned as information identifying this precoding (e.g., the PMI indicator) and information indicating the number of spatial layers (e.g., the RI indicator). This combination (e.g., (PMI, RI)) can be the one that maximizes the receive throughput.

[0242] A selection of the best v+k beams from the beam dictionary based on a first performance criterion related to receiver power has been described above. However, other first performance criteria can be considered, such as the SINR for a beam, or a combination of the best SINR and the best receiver power as described above. The best matrix(s) according to the second performance criterion are then chosen from among these best v+k beams according to this SINR. Furthermore, the determination of the best v+k beams can be carried out within a particular subset of the beam dictionary.

[0243] An embodiment of the invention has been described above in which the second performance criterion, on the basis of which the precoding matrix to be transmitted is determined, is a receive throughput for a given target error rate (BLER) of 10% (for an eMBB service), i.e., the predicted BLER must be less than or equal to this target BLER. For a so-called Ultra Reliable Low Latency service, this BLER can be 10 -5 However, the invention also relates to other secondary performance criteria, for example, the selection of a precoding matrix with the best effective SINR. Other performance indicators can be considered, such as a quality of service (QoS) indicator like maximizing throughput for a bit error rate corresponding to the target QoS. Other QoS indicators can be considered such as BLER, effective SINR, delay / latency, etc.

[0244] 5.4. Devices

[0245] Finally, in relation to figures 3 and 4, the simplified structures of a receiver and a transmitter are presented according to an example of an embodiment of the invention.

[0246] As illustrated in Figure 3, a receiver 3 according to one embodiment of the invention comprises a memory M, a processing unit P, equipped for example with a programmable computing machine or a dedicated computing machine, for example a processor, and controlled by a computer program Pg, implementing steps of the process described above.

[0247] At initialization, the code instructions of the computer program Pg are, for example, loaded into RAM memory before being executed by the processor of the processing unit P.

[0248] The processor of the processing unit P implements steps of the communication process described above, according to the instructions of the computer program Pg, to:

[0249] select, from a dictionary of beams (B), a number v + k of beams exhibiting the best performance with respect to a first performance criterion, with k a strictly positive integer;

[0250] determine at least two broadband precoding matrices (W1 power W1 inter ) distinct, each formed from v beams among said v + k selected beams; and transmit, to said transmitter, at least one information (PMI) identifying the broadband precoding matrix, among said determined broadband precoding matrices, defining a precoding (W) exhibiting better performance with respect to a second performance criterion, compared with at least one precoding defined by another of said determined broadband precoding matrices.

[0251] As illustrated in Figure 4, a transmitter according to one embodiment of the invention comprises a memory M', a processing unit P', equipped, for example, with a programmable computing machine or a dedicated computing machine, for example a processor, and controlled by a computer program Pg', implementing steps of the process described above. At initialization, the code instructions of the computer program Pg' are, for example, loaded into RAM before being executed by the processor of the processing unit P'.

[0252] The processor of the processing unit P' implements steps of the communication process described above, according to the instructions of the computer program Pg', to: - receive, from the receiver, at least a second identifier associated with a second broadband precoding matrix W1interference minimizing beam interference, the second broadband precoding matrix being determined on the basis of a set of v + k beams included in a beam dictionary (B), the beams of said set of v + k beams exhibiting the best performance with respect to a first performance criterion, with k a strictly positive integer; and

[0253] implement precoding based on a precoding matrix that is a function of at least one identifier associated with the second broadband precoding matrix.

Claims

DEMANDS 1. A method for determining, by a receiver, at least one precoding matrix, for implementing a transmission from a transmitter to said receiver through a transmission channel, said method comprising, for at least a given number of spatial layers v: the selection (E1), within a dictionary of beams (B), of a number v + k of beams with k a strictly positive integer; and the determination (E2) of at least two broadband precoding matrices ( W1 power W1 inter ) distinct, each formed from v beams among said v + k selected beams.

2. Method of determination according to claim 1 in which the v + k selected beams are those which have the best performance within the beam dictionary (B) with regard to a first performance criterion.

3. Method of determination according to claim 2, wherein the first performance criterion, on the basis of which the v + k beams are selected in the beam dictionary, is a power in reception at the receiver.

4. Method of determination according to any one of claims 1 to 3 further comprising the transmission (E3), to said transmitter, of at least one piece of information (PMI) identifying the broadband precoding matrix, among said determined broadband precoding matrices, defining a precoding (W) exhibiting better performance with regard to a second performance criterion, compared with at least one precoding defined by another of said determined broadband precoding matrices.

5. Method of determination according to claim 4, wherein the second performance criterion, on the basis of which information (PMI) representative of a precoding matrix is ​​transmitted, is a receive rate.

6. A method for determining according to any one of claims 2 and 3 to 5 when they depend on claim 2, wherein said at least two broadband precoding matrices comprise at least one broadband precoding matrix ( W1 power ) formed from the v beams exhibiting the best performance with regard to said first performance criterion.

7. Method of determination according to claim 6, wherein said determination of a plurality of precoding matrices is carried out for all distinct broadband precoding matrices formed from the selection of v beams from said v + k beams, and wherein the broadband precoding matrix whose identifier is transmitted to the transmitter is the one whose resulting precoding has a better second performance criterion than that of any other precoding formed by any of the other broadband precoding matrices.

8. A method for determining according to claim 6 or 7, wherein said determination of a plurality of precoding matrices is carried out for two distinct broadband precoding matrices formed from the selection of v beams from said v + k beams, comprising: a first broadband precoding matrix corresponding to said broadband precoding matrix (W1 power ) formed from the v beams associated with better reception power at said receiver; a second broadband precoding matrix (W1 inter ), formed from the v beams minimizing interference between spatial layers.

9. A method for determining according to claim 8, wherein said second broadband precoding matrix (W1 inter ) is determined to be the broadband precoding matrix (VK^), among all distinct broadband precoding matrices formed from the selection of v beams from said v + k beams, minimizing interference defined by the following formula: with p = 1 or 2 mtn where | | p represents the norm p, b n (i)represents the i-th coefficient of the beam b n taken from said v + k bundles, and R n and R22 are the diagonal blocks of the R interference covariance matrix.

10. A determination method according to claim 8 or 9, wherein said transmission to said transmitter comprises the joint transmission of an identifier associated with the first broadband precoding matrix (W1 power ) and an identifier associated with the second broadband precoding matrix (W1 inter ).

11. Method of determination according to claim 10, wherein said joint transmission is conditioned on a prior verification of a configuration data previously received from said transmitter.

12. A method of determination according to any one of the preceding claims, wherein said number k is data received from said sender, prior to said selection.

13. A method for determining according to any one of the preceding claims, wherein said number k is of the same order of magnitude as a maximum number of spatial layers (y max ).

14. Communication method for implementing a transmission from a transmitter to a receiver through a transmission channel, said method being implemented by the transmitter and comprising, for at least a given number of spatial layers v: a reception, from the receiver, of at least a second identifier associated with a second broadband precoding matrix (W1 inter) minimizing beam interference, the second broadband precoding matrix being determined on the basis of a set of v + k beams included in a beam dictionary (B), with k a strictly positive integer; and an implementation of precoding on the basis of a precoding matrix function of at least one identifier associated with the second broadband precoding matrix.

15. Communication method according to claim 14 wherein the beams of said set of v + k beams are those which exhibit the best performance within the beam dictionary (B) with regard to a first performance criterion.

16. A communication method according to claim 14 or 15, further comprising, in conjunction with said reception of at least a second identifier associated with the second matrix, a reception, from the same receiver, of a first identifier associated with a first broadband precoding matrix (W1 power ) determined on the basis of the v bundles of the dictionary (B) which maximize the first performance criterion.

17. Communication method according to claim 16, further comprising prior to the reception of at least a second identifier, the transmission of a control message comprising configuration data requiring a joint transmission of said identifiers of said first and second broadband precoding matrices.

18. Receiver configured to determine at least one precoding matrix for implementing a transmission from a transmitter to said receiver through a transmission channel, said receiver comprising at least one processor configured for at least a given number of spatial layers v: select (E1), within a dictionary of sheaves (B), a number v + k of sheaves, with k a strictly positive integer; and determine (E2) at least two broadband precoding matrices ( W1 power W1 inter ) distinct, each formed from v beams among said v + k selected beams.

19. Transmitter configured for the implementation of a transmission from said transmitter to a receiver through a transmission channel, the transmitter comprising at least one processor configured for, for at least a given number of spatial layers v: receive, from the receiver, at least a second identifier associated with a second broadband precoding matrix (i ” lter / erence ) minimizing beam interference, the second broadband precoding matrix being determined on the basis of a set of v + k beams included in a beam dictionary (B), with k a strictly positive integer; and implement precoding based on a precoding matrix that is a function of at least one identifier associated with the second broadband precoding matrix.

20. Product computer program comprising instructions for the implementation of a method according to any one of claims 1 to 17 when this program is executed by a processor.

Images