OMAMRC METHOD AND SYSTEM WITH FDM TRANSMISSION

Abstract

A message method intended for an OMAMRC telecommunication system with M sources s.sub.i iε{1, . . . , M], potentially L relays (r.sub.1 . . . , r.sub.L) and a destination. The transmission is of FDM type in a band divided into B mutually orthogonal sub-bands. The method includes: a simultaneous transmission of the M sources in a time interval with allocation of at least one sub-band per source and at least one cooperative retransmission for a time interval of at least one relay node taken from among the M sources and the L relays which is selected using a selection strategy with allocation of at least one sub-band per node selected.

Claims

1. A transmission method comprising: transmitting frame messages intended for a telecommunication system with a transmission channel, M sources s.sub.iiϵ{1, . . . , M}, optionally L relays (r.sub.1, . . . , r.sub.L) and a destination (d), M≥2, L≥0, M≤B and orthogonal multiple-access to the transmission channel, wherein the transmitting is of a frequency-division multiplexing (FDM) type on a band divided into B mutually orthogonal sub-bands, and wherein the transmitting comprises: simultaneous transmission of the M sources during a time slot with allocation of at least one sub-band per source; and at least one cooperative retransmission of a time slot of at least one relay node taken from among the M sources and the L relays selected by the destination, with allocation by the destination of at least one sub-band per selected node, by knowing the sources correctly decoded by the nodes the destination selects the nodes that allow as many newly correctly decoded sources as possible to be obtained at the destination on completion of the cooperative retransmission.

2. The transmission method as claimed in claim 1, wherein the selection by the destination is such that a relay node that decodes a set of sources at a time slot t can only cooperate at a time slot t+1 for a single source of its set.

3. The transmission method as claimed in claim 1, wherein: the destination broadcasts its set of correctly decoded sources from among the received sources to the relay nodes during a transmission slot; the relay nodes that have correctly decoded a source not correctly decoded by the destination notify the destination as such; the destination broadcasts a vector to the relay nodes comprising the relay nodes selected for the sub-bands for cooperative or non-cooperative retransmission during the next transmission slot.

4. The transmission method as claimed in claim 3, wherein a relay node notifies the destination by transmitting its set of correctly decoded sources.

5. The transmission method as claimed in claim 1, wherein the destination selects the relay nodes such that the sum of the mutual information between the nodes that can assist with their allocated sub-bands and the destination is maximized.

6. The transmission method as claimed in claim 1, further comprising an initial phase of determining initial rates and such that the initial rates allocated to the sources are determined in order to maximize a metric expressed in the form of an average utility function subject to an average individual BLER for each source: ${\eta }^{sla}=\frac{1}{B}{.Math.}_{i=1}^M\frac{R_i{n}_{0,i}}{1+𝔼\left(T_{used}\right)}\left(1-BLE{R}_i\right),$ with $R_i=\frac{K_i}{n_{0,i}\times F}$ being a variable representing the initial rate allocated to the source i, i∈{1, . . . , M}; K.sub.i being the amount of data transmitted on n.sub.0,i×F channel uses by the source i; T.sub.used being the number of time slots used for cooperative or non-cooperative retransmissions; (T.sub.used) being an average of the number of time slots used for the cooperative or non-cooperative retransmission; BLER.sub.i being the block error rate for the source i.

7. The transmission method as claimed in claim 1, further comprising an initial phase of determining initial rates at each frame and such that the initial rates allocated to the sources are determined in order to maximize a metric expressed in the form of an average utility function subject to individual cutoffs of the sources: ${\eta }^{\mathrm{fla}}=\frac{1}{B}{.Math.}_{i=1}^M\frac{R_i{n}_{0,i}}{1+T_{used}}\left\{1-{𝒪}_{i,T_{used}}\right\},$ with: .sub.i,t being the individual cutoff probability of the source i at the cooperative or non-cooperative retransmission slot t; T.sub.used being the number of cooperative or non-cooperative retransmissions; $R_i=\frac{K_i}{n_{0,i}\times F}$ being a variable representing the initial rate allocated to the source i, i∈{1, . . . , M}; n.sub.0,i being a number of sub-bands allocated to the node i for the time slot 0, i∈{1, . . . , M}; F being a number of channel uses.

8. A telecommunication system comprising: M sources (s.sub.1, . . . , s.sub.M), L relays (r.sub.1, . . . , r.sub.L); and a destination (d), M≥2, L≥0, wherein the M sources each comprise a processor configured to transmit frame messages intended for the telecommunication system with a transmission channel and orthogonal multiple-access to the transmission channel, wherein the transmitting is of a frequency-division multiplexing (FDM) type on a band divided into B mutually orthogonal sub-bands, and wherein the transmitting comprises: simultaneous transmission of the M sources during a time slot with allocation of at least one sub-band per source; and at least one cooperative retransmission of a time slot of at least one relay node taken from among the M sources and the L relays selected by the destination, with allocation by the destination of at least one sub-band per selected node, by knowing the sources correctly decoded by the nodes the destination selects the nodes that allow as many newly correctly decoded sources as possible to be obtained at the destination on completion of the cooperative retransmission.

Description

LIST OF FIGURES

[0057] Further features and advantages of the invention will become more clearly apparent from reading the following description of embodiments, which are provided by way of simple illustrative and non-limiting examples, and the appended drawings, in which:

[0058] FIG. 1 is a diagram of an example of an OMAMRC (Orthogonal Multiple-Access Multiple Relays Channel) system according to the invention;

[0059] FIG. 2 is a diagram of a transmission cycle of a frame according to one embodiment of the invention;

[0060] FIG. 3 is a diagram of the protocol of the exchanges of information between the destination and the nodes, sources and relays, according to one embodiment of the invention.

DESCRIPTION OF SPECIFIC EMBODIMENTS

[0061] A channel use is the smallest time-frequency resource grading defined by the system that allows a modulated symbol to be transmitted. The number of channel uses is linked to the available frequency band and to the transmission duration.

[0062] An OMAMRC system is illustrated in FIG. 1. Such a system comprises M sources that belong to the set of sources ={1, . . . , M}, L relays that belong to the set of relays ={M+1, . . . , M+L} and a destination d.

[0063] Each source of the set communicates with the single destination with the assistance of the other sources (user cooperation) an relays that cooperate.

[0064] By way of a simplification of the description, the following assumptions are made hereafter with respect to the OMAMRC system: [0065] the sources, the relays are equipped with a single transmission antenna; [0066] the sources, the relays, and the destination are equipped with a single reception antenna; [0067] the sources, the relays, and the destination are perfectly synchronized; [0068] the sources are statistically independent (there is no correlation between them); [0069] all the nodes transmit with the same power; [0070] a CRC code is used that is assumed to be included in the K.sub.i information bits of each source i in order to determine whether or not a message is correctly decoded, i∈; [0071] the links between the various nodes experience additive noise and fading. The fading gains are fixed when a frame is transmitted over a maximum duration of 1+T.sub.max time slots, but can change independently from one frame to another. T.sub.max≥1 is a parameter of the system; [0072] the instantaneous quality of the direct reception channel/link (CSIR “Channel State Information at Receiver”) is available at the destination, the sources and the relays; [0073] feedback is error-free (no error on the control signals).

[0074] The nodes comprise the relays and the sources that can act as a relay when they do not emit their own message.

[0075] The nodes, M sources and L relays, access the transmission channel according to a frequency orthogonal multiple-access scheme and operate in accordance with a full-duplex mode that allows them to listen to the transmissions of the other nodes without any interference.

[0076] The band of the channel is divided into B sub-bands, the number of which is assumed to be greater than or equal to the number of sources: B≥M. Each sub-band associated with a time slot determines F channel uses (F resource elements).

[0077] In the case of a transmission with OFDM modulation, a sub-band can comprise, for example, as many sub-carriers as an OFDM symbol.

[0078] The number N of channel uses is assumed to be identical for each transmission slot: N=B ×F. A transmission cycle lasts for 1+T.sub.used time slots, with T.sub.used≤T.sub.max and T.sub.max being the maximum number of time slots. At each time slot, no sub-band, or one or more sub-band(s) is/are allocated to a node according to a first partition.

[0079] During the first time slot (first phase) all the sources transmit, assuming that B≥M, respectively on one or more sub-band(s) allocated to each source.

[0080] During the next “retransmission slots” (second phase), only the nodes selected from among the sources and the relays retransmit and their retransmission occurs on the one or more sub-band(s) that are respectively allocated thereto according to a partition determined for each current slot. Thus, the partitions can differ between all the transmission slots, including the first.

[0081] The selection of the nodes and the allocation of the sub-bands are implemented by a scheduler, typically hosted by the destination.

[0082] The following notations are used: [0083] if i≤M the selected node i is a source i denoted s.sub.i, i∈{1, . . . , M}, otherwise i>M and the selected node is a relay i−M denoted r.sub.i-M, i∈{M+1, . . . , M+L}; [0084] a.sub.t∈(∪).sup.B is the dimension vector B of the nodes selected for the transmission slot t, whether this is during the first phase or during the second phase. The i.sup.th element a.sub.t,i of the vector a.sub.t designates the i.sup.th sub-band and the active selected node during this time slot t in this sub-band i, i∈{1, . . . , B}. The order in the vector corresponds to the order of the sub-bands; [0085] n.sub.t∈{0, . . . , B}.sup.M+L is the dimension vector M+L of the number of sub-bands allocated for each node that varies between 0 (the node is inactive) and B (the node occupies all the sub-bands), source or relay, for the transmission slot (time slot) t, whether this is during the first phase or during the second phase. The i.sup.th element n.sub.t,i of the vector n.sub.t denotes the number of sub-bands allocated to the node i at the transmission slot (time slot) t, i∈{1, . . . , M+L}. The sum of the elements forming the vector n.sub.t is equal to B, the number of sub-bands; [0086] h.sub.a,b is the attenuation gain of the channel (fading) between the node a (source or relay) and the node b (source, relay or destination) that follows a symmetrical circular complex Gaussian distribution with a zero average and variance γ.sub.a,b the gains are mutually independent;

[0087] T.sub.used is the minimum number of retransmission time slots, i.e., during the second phase that leads to zero faults for all the sources (the individual cutoff event of each of the sources is zero):

[00006]$\begin{array}{cc}{T}_{used}=\underset{t\le T_{\max }}{\min }ts.t.\left({i,t}_{}=0\forall i\mathrm{or}t=T_{\max }\right)& \left(1\right)\end{array}$

[0088] The individual cutoff event .sub.s,t(a.sub.t,n.sub.t, .sub.a.sub.t.sub.,t-1|h.sub.dir,.sub.t-1) of the source s after the retransmission slot t (round t) depends on the vector a.sub.t for selecting the nodes, the vector n.sub.t for allocating sub-bands and the set .sub.a.sub.t.sub.,t-1 of decoded sources at the end of the preceding slot, t−1.

[0089] It is also contingent upon knowledge of the implementations of the channel of the direct links h.sub.dir (the gains of the channel), as well as on .sub.t-1. .sub.t-1 designates the set of selection vectors â.sub.l (therefore, of selected nodes) and of allocation vectors {circumflex over (n)}.sub.l with their associated set of decoded sources .sub.â.sub.l.sub.,l-1 determined for the slots (rounds) l preceding the slot t, l∈{1, . . . , t−1} and the set .sub.d,t-1 of sources decoded by the destination. It should be noted that a.sub.0 is the selection vector of the source nodes transmitting during the transmission phase, that n.sub.0 is the allocation vector of sub-bands allocated for each source during the transmission phase and that .sub.d,0 is the set of sources decoded by the destination on completion of the first phase.

[0090] The common cutoff event (a.sub.t, n.sub.t,.sub.a.sub.t.sub.,t−1|h.sub.dir,.sub.t-1) for the sub-set of sources after the time slot t (round t) is the event whereby at least one source of the sub-set is not correctly decoded by the destination at the end of this slot t. Subsequently, the dependencies of .sub.d,t-1 with h.sub.dir and with .sub.t-1 are omitted in order to simplify the notations. .sub.d,t=\.sub.d,t denotes the set of sources not successfully decoded by the destination at the end of the time slot t (round t).

[0091] From an analytical perspective, the common cutoff event of a sub-set of sources occurs, i.e., is met, if the vector of the rates of these sources is not included in the corresponding capacity region MAC.

[0092] Thus, for a given sub-set of sources .Math..sub.d,t-1, for a candidate vector a.sub.t of selected nodes and the corresponding sub-band allocation vector n.sub.t, this event can be expressed in the following form:

(a.sub.t,n.sub.t,.sub.a.sub.it.sub.,t-1)=(),  (2)

where () expresses the non-compliance of the inequality MAC associated with the sum rate of the sources contained in :

()=[R.sub.sn.sub.0,s>Ī.sub.0,s+Σ.sub.l=1.sup.t−1Σ.sub.i=1.sup.M+LĪ.sub.l,i[.sub.l,i]+Σ.sub.i=1.sup.M+LĪ.sub.t,i[.sub.t,i]]  (3)

with: [0093] l being the time slot index (round) of the second phase with the convention that l=0 corresponds to the end of the first phase (transmission phase), l∈{1, . . . , T.sub.used}; [0094] s being the index corresponding to the source node, s∈{1, . . . , M}; [0095] i being the index corresponding to any node (source and relay), i∈{1, . . . , M+L}; [0096] n.sub.l,i being the number of sub-bands allocated to the node i for the time slot l (round) l∈{1, . . . , T.sub.used}; [0097] n.sub.0,s being the number of sub-bands allocated to the source s∈{1, . . . , M} by the destination for the first phase;

.sub.l,i=[[.sub.i,l−1∩≠∅]∧[.sub.i,l−1∩=∅]]  (4)  with =.sub.d,t−1\, representing the set of interfering sources, .sub.l,i equals one if, on the one hand, the intersection between the set of sources correctly decoded by the node i at the slot l−1 and the set is not empty and, on the other hand, the intersection between the set of sources correctly decoded by the node i at the slot l−1 and the set of interfering sources is empty; [0098] ∧ represents the “and” logic; [0099] [P] represents the Iverson brackets, i.e., that yield the value 1 if the event P is met and the value 0 if not; [0100] Ī.sub.i,l being the block of mutual fading information of the node i at the destination d for the sub-bands n.sub.l,i allocated to the node i at the time slot l∈{1, . . . , T.sub.used}:

Ī.sub.l,i=Σ.sub.f=1.sup.BI.sub.a.sub.l,f.sub.,d[i=a.sub.l,f]  (5)  where I.sub.a.sub.l,f.sub.,d is the mutual information between the node a.sub.l,f to which the sub-band f is allocated at the time slot (round) l∈{1, . . . , T.sub.used} and the destination d. The mutual information depends on the power transmitted on the sub-band of the channel, i.e.,

[00007]$\frac{P_T}{n_{l,a_{l,f}}}$

between the node a.sub.l,f and the destination d, with P.sub.T being the total power of this node. If the node i is not selected at the time slot I, then the mutual information block Ī.sub.l,i is zero; [0101] Ī.sub.0,s being the block of mutual fading information of the source s at the destination d, for given a.sub.0 and n.sub.0, at the time slot corresponding to the transmission phase (first phase); [0102] R.sub.s=K.sub.s/(n.sub.0,sF), s=1, . . . , M is the rate used during the first phase, with K.sub.s being the number of useful information bits transmitted on n.sub.0,sF channel uses.

[0103] Subsequently, the cutoff event for a given source s is defined in the following form:

[00008]${s,t}_{}\left(a_t,n_t,{a_t,t-1}_{}\right)=\underset{𝒥\subset {\stackrel{\_}{𝒮}}_{d,t-1},ℬ=\stackrel{\_}{𝒥},s\in ℬ}{.Math.}{ℰ}_{t,ℬ}\left(a_t,n_t,{a_t,t-1}_{}\right)$

which by definition is the intersection of all the common cutoff events corresponding to a set of sources including the source s. A source s is cutoff if, and only if, there is no set of sources including it that can be associated with error-free decoding, i.e., =0. It becomes:

[00009]${s,t}_{}\left(a_t,n_t,{a_t,t-1}_{}\right)=\underset{𝒥\subset {\stackrel{\_}{𝒮}}_{d,t-1}}{.Math.}\underset{𝒰.Math.\stackrel{\_}{𝒥}:s\in \stackrel{\_}{𝒥}}{.Math.}\left\{\underset{i\in 𝒰}{.Math.}{R}_i{n}_{0,i}>\underset{i\in 𝒰}{.Math.}{\stackrel{\_}{I}}_{0,i}+\underset{l=1}{\overset{t-1}{.Math.}}\underset{i=1}{\overset{M+L}{.Math.}}{\stackrel{\_}{I}}_{l,i}{1}_{\left\{{𝒞}_{l,i}\right\}}+\underset{i=1}{\overset{M+L}{.Math.}}{\stackrel{\_}{I}}_{t,i}{1}_{\left\{{𝒞}_{t,i}\right\}}\right\}.$

[0104] This cutoff event indicates whether a source is decoded without error (.sub.s,t=0) or if it is cutoff (.sub.s,t=1). This approach allows the result of the implementation of a parity check (CRC check) to be predicted without proceeding with the simulation of the whole of the transmission (modulation coding) and reception (detection/demodulation, decoding) chain. In this way, it defines an abstraction of the physical layer. Some adjustments obtained by simulation (called calibration within the context of abstractions of the physical layer) for a given coding scheme can be carried out by introducing weighting parameters of the mutual information and/or of the SNR of the links.

[0105] The two transmission phases of the transmission method can be preceded by an initial phase of determining an initial rate. This phase can occur once every several hundred frames (i.e., each time the quality statistics of the channel/link change) in the case of slow fading, this is referred to as slow link adaptation. Alternatively, this phase can occur much more frequently and at most at each cycle, this is referred to as fast link adaptation. Whether the link adaptation is fast or slow, the rate of each source and the allocation of sub-bands are known before the start of the transmission.

[0106] By exploiting reference signals (pilot symbols of the 3GPP LTE/NR DMRS type, reference signals of the 3GPP LTE/NR SRS type, etc.), the destination can determine the gains (CSI “Channel State Information”) of the direct links: h.sub.dir={h.sub.s.sub.1.sub.,d, . . . , h.sub.s.sub.M.sub.,d, h.sub.r.sub.1.sub.,d, . . . , h.sub.r.sub.L.sub.,d}, i.e., source links to destination and relay to destination. The destination can therefore deduce average values therefrom for the direct links within the context of slow adaptation.

[0107] By contrast, the gains of the links between sources, of the links between relays and of the links between sources and relays are not known to the destination. Only the sources and the relays can estimate a metric of these links by exploiting reference signals in a manner similar to that used for the direct links.

[0108] Given that within the context of slow adaptation, the statistics of the channels are assumed to be constant between two initialization phases, transmitting metrics to the destination by the sources and the relays can only occur at the same rate as the initialization phase. The statistic of the channel of each link is assumed to follow a centered circular complex Gaussian distribution and the statistics are independent between the links.

[0109] Within the scope of fast adaptation, optimizing the spectral efficiency can be based on knowledge of all the links or of some links. One possible, but very onerous solution in terms of control, is that the sources and the relays feed back the coefficients of the links (quantified per sub-band) that they can estimate to the destination.

[0110] During the initial link adaptation phase that precedes the transmission of one or more frame(s), the destination transmits, for each source s, a representative value (index, MCS, rate, etc.) of an initial rate R.sub.i. Each of the initial rates unambiguously determines an initial modulation and coding scheme (MCS) or, conversely, each initial MCS determines an initial rate. The initial rates R.sub.i are fed back via control channels with a very limited rate.

[0111] These initial rates are determined by the destination so as to maximize a quality of service metric, for example, an average spectral efficiency.

[0112] In the case of slow link adaptation, the quality of service metric is, according to one embodiment, an average spectral efficiency, which is expressed in the following form:

[00010]$\begin{array}{cc}{\eta }^{sla}=\frac{1}{B}{.Math.}_{i=1}^M\frac{R_i{n}_{0,i}}{1+𝔼\left(T_{used}\right)}\left(1-BLE{R}_i\right),& \left(6\right)\end{array}$

with:

[00011]$R_i=\frac{K_i}{n_{0,i}\times F}$

being a variable representing the initial rate allocated to the source i, i∈{1, . . . , M}; [0113] K.sub.i being the amount of data transmitted on n.sub.0,i×F channel uses by the source i; [0114] T.sub.used being number of time slots used for the cooperative or non-cooperative retransmissions; [0115] (T.sub.used being an average of the number of time slots used for the retransmissions, whether they are cooperative or non-cooperative; [0116] BLER.sub.i being the average block error rate for the source i.

[0117] In the case of slow link adaptation, the rate and the allocation of sub-bands per source remains unchanged for several hundred transmissions of messages from the sources, which allows the block error rate (BLER) of the source i to be averaged on the statistics of the channel (CDI “Channel Distribution Information”) known to the destination. A source i takes n.sub.0,i×F resource elements in order to transmit the data K.sub.i at the rate R.sub.i during the time slot of the first phase.

[0118] In the case of fast link adaptation, the quality of service metric is, according to one embodiment, an average utility function subject to the individual cutoffs of the sources defined per transmitted message and the rate and the allocation of sub-bands can change from one message to the next:

[00012]$\begin{array}{cc}{\eta }^{fla}=\frac{1}{B}{.Math.}_{i=1}^M\frac{R_i{n}_{0,i}}{1+T_{used}}\left(1-{𝒪}_{i,T_{used}}\right)& \left(7\right)\end{array}$

with: [0119] .sub.i,t being the individual cutoff event of the source i at the retransmission slot t, which equals one in the event of a fault or zero in the event of success (correctly decoded source), [0120] .sub.i,0 is the individual cutoff event at the end of the transmission phase (first phase of a time slot); [0121] T.sub.used being the number of time slots used for cooperative or non-cooperative retransmissions (the second phase can assume the value 0 if .sub.i,0=0∀i∈);

[00013]$R_i=\frac{K_i}{n_{0,i}\times F}$

being a variable representing the initial rate allocated to the source i, i∈{1, . . . , M}.

[0122] One embodiment of a transmission method according to the invention is described with reference to the diagram of FIG. 2, which illustrates a transmission cycle of a frame within the context of a three-source OMAMRC system, i=(1, 2, 3) two relays, i=(4, 5), a destination and a transmission channel with a certain bandwidth. The band is split into B=5 sub-bands and each sub-band associated with a time slot determines F=5 channel uses, that is, N=25.

[0123] During the first phase of a time slot, each source i={1, 2, 3) emits its code words. According to the example, the number of sub-bands allocated to a source differs between the sources. Thus, the sub-bands f.sub.1,f.sub.2 and f.sub.5 are allocated to the source 1, the sub-band f.sub.3 is allocated to the source 2 and the sub-band f.sub.4 is allocated to the source 3. The selection vector is therefore a.sub.0=[s.sub.1,s.sub.1,s.sub.2,s.sub.3,s.sub.1]T=[1,1,2,3,1].sup.T. The vector for allocating sub-bands per node is therefore n.sub.0=[3,1,1,0,0].sup.T.

[0124] During the second phase, called retransmission phase, and for the first time slot, only the sources 2, 3 and the relay 2 are selected and the sub-band f.sub.1 is allocated to the source 3, the sub-bands f.sub.2,f.sub.3 and f.sub.4 are allocated to the relay 5 and the sub-band f.sub.5 is allocated to the source 2. The selection vector is therefore a.sub.1=[s.sub.3,r.sub.2,r.sub.2,r.sub.2,s.sub.2].sup.T=[3,5,5,5,2].sup.T. The vector for allocating sub-bands per node is therefore n.sub.1=[0,1,1,0,3].sup.T.

[0125] During the second phase, called retransmission phase, and for the second time slot, only the source 1 and the relay 4 are selected and the sub-bands f.sub.1, f.sub.2 and f.sub.5 are allocated to the relay 4, the sub-bands f.sub.3 and f.sub.4 are allocated to the source 1. The selection vector is therefore a.sub.2=[r.sub.1,r.sub.1,s.sub.1,s.sub.1,r.sub.1].sup.T=[4,4,1,1,4].sup.T. The vector for allocating sub-bands per node is therefore n.sub.2=[2,0,0,3,0].sup.T.

[0126] An embodiment of the protocol of the exchanges between the nodes and the destination is illustrated in FIG. 3.

[0127] Each source transmits its frame data to the destination with the assistance of the other sources and of the relays. A frame occupies time slots during the transmission of the M messages of the M sources, respectively. The transmission of a frame (which defines a transmission cycle) occurs for 1+T.sub.used time slots: 1 slot for the 1.sup.st phase with a channel uses capacity n.sub.0,i for each source i, T.sub.used slots for the 2.sup.nd phase with a channel uses capacity n.sub.t,i for each source i.

[0128] During the first phase, each source s∈ transmits, after encoding, a message u.sub.s comprising K.sub.s information bits u.sub.s∈.sub.2.sup.K.sup.s, with .sub.2 being the two element Galois field. The message u.sub.s comprises a code of the CRC type, which allows the integrity of the message u.sub.s to be checked. The message u.sub.s is coded according to the initial MCS. Given that the initial MCSs can be different between the sources, the lengths of the coded messages can be different between the sources. The coding uses an incremental redundancy code. The code word that is obtained is segmented into redundancy blocks. The incremental redundancy code can be of the systematic type, the information bits are then included in the first block. Whether or not the incremental redundancy code is of the systematic type, it is such that the first block can be decoded independently of the other blocks. The incremental redundancy code can be produced, for example, by means of a finite family of rate compatible punctured linear codes or of rate-free codes modified in order to operate with finite lengths: raptor code (RC), rate compatible punctured turbo code (RCPTC), rate compatible punctured convolutional code (RCPCC), rate compatible low density parity check code (RCLDPC).

[0129] Thus, during the first phase, the M sources simultaneously transmit their message during the transmission slot on the allocated sub-hands in accordance with the vector a.sub.0, respectively with modulation and coding schemes determined from the values of the initial rates.

[0130] Each transmitted message corresponding to a source s∈, a correctly decoded message, is assimilated with the corresponding source for the purposes of notation.

[0131] Whether it is during the first phase or the second phase, when a node, in particular a source, transmits, the destination and the other nodes listen. Each full-duplex node can simultaneously transmit and listen to all the other nodes given that, in order to transmit, each node is allocated one or more sub-band(s) that are different between the nodes.

[0132] The destination, the sources and the relays attempt to decode the messages received at the end of a time slot. The success of the decoding at each node is determined using the CRC. The destination and the nodes thus determine their set of correctly decoded sources.

[0133] During the second phase, at the time lot (round) t, the destination d transmits its set of correctly decoded sources at the end of the preceding time slot .sub.d,t−1 by using, for example, a feedback broadcast control channel t={1, . . . , T.sub.used}. This feedback can be made up of a vector of M bits.

[0134] If the decoding of all the sources by the destination is correct, .sub.d,t−i=S. In this case, the current cycle is stopped, a new cycle can start. A cycle for transmitting a new frame begins with deleting the memories of the relays and of the destination and with the transmission of new messages by the sources. The number of time slots (rounds) used during the second phase T.sub.used={1, . . . , T.sub.max} depends on the success of decoding at the destination.

[0135] The nodes, sources and relays compare the set .sub.d,t−1 with their set of correctly decoded sources.

[0136] If the set of a node comprises at least one source not included in the set .sub.d,t−1 of the destination, the node notifies the destination accordingly by using, for example, a dedicated control channel of the unicast type. The information transmitted by a node can be made up of its set of correctly decoded sources or, as illustrated in FIG. 3, a bit, for example, set to one.

[0137] During this second phase, the destination follows a certain strategy in order to determine the one or more selected node(s) that transmit at the time slot (round) t.

[0138] The destination notifies the nodes of this selection by transmitting the vector a.sub.t using, for example, the feedback broadcast control channel.

[0139] Each node that receives the vector a.sub.t can determine whether it is selected and on which sub-band it is to transmit.

[0140] During this second phase, and for at least one retransmission slot from among the T.sub.used retransmission slots, at least one selected node, source or relay generates a cooperative retransmission. Outside the at least one time slot, the retransmissions can be cooperative or non-cooperative.

[0141] The node selected for a retransmission transmits u.sub.a.sub.t, after multi-user coding, the words or some of the words that it has correctly decoded. The selected node can transmit parities determined on the basis of the messages of its set of correctly decoded sources using network coding and Joint Network Channel Coding. The other nodes and the destination can improve their own decoding by exploiting the transmission of the selected node and can consequently update their set of correctly decoded sources.

[0142] The destination thus controls the transmission of the nodes using a feedback channel. This allows the spectral efficiency and the reliability to be improved by increasing the probability of decoding all the sources by the destination.

Selection Strategy

[0143] According to a first strategy, the destination selects those that maximize the sum of the mutual information from among the set Zt of the various allocations of sub-bands to the nodes that can assist at the slot t. This strategy only requires knowledge of the nodes that can assist, it is compatible with the mode in which the nodes transmit information in the form of a bit.

[0144] The selection criterion can be expressed in the following form:

â.sub.t=argmax.sub.a.sub.t.sub.∈Z.sub.t{Σ.sub.i=1.sup.M+LĪ.sub.t,i}  (8)

with Z.sub.t being the set of possible vectors a.sub.t that correspond to the selection of the nodes that can assist the destination at the slot (round) t.

[0145] On the basis of the sets of correctly decoded sources received from the nodes and according to a second strategy, the destination selects the nodes that allow as many newly correctly decoded sources as possible to be obtained at the destination at the end of the current slot t, i.e., which maximize the cardinality of the set of sources correctly decoded by the destination at the end of the current slot t.

[0146] According to this strategy, the method reviews all the possible values of the vector a.sub.t and retains that which leads to the greatest number of newly decoded sources. Thus, the method does not take into account the nodes that cannot assist the sources not yet decoded, since it targets the greatest number of newly decoded sources, i.e., the only nodes i that are considered are those that meet: .sub.d,t−1∩.sub.a.sub.t,i.sub.,t−1≠∅ for i∈{1, . . . , M+L}.

[0147] This strategy also requires knowledge of the set of correctly decoded sources of all the previously selected nodes.

[0148] In the case whereby several vectors a.sub.t can lead to the same maximum number of newly decoded sources, the method selects the vector â.sub.t that maximizes the sum Σ.sub.i=1.sup.M+LĪ.sup.t,i of the mutual information. Indeed, at a time slot t, this is the only element that can be maximized in order to maximize the straight part of the individual cutoff events and common cutoff events. The presence of .sub.t,i in the expression of the common cutoff event expresses the fact that the only nodes that can be selected are those that can assist, i.e., have decoded at least one source not yet decoded by the destination. The selection criterion then can be expressed in the following form:

â.sub.t=argmax.sub.a.sub.t.sub.∈{dot over (A)}.sub.t{Σ.sub.i=1.sup.M+LĪ.sub.t,i}  (9)

with {dot over (A)}.sub.t being the set of candidate nodes that maximize the set of the destination at the end of the slot (round) t.

[0149] It should be noted that for t=0, the only candidate nodes for the first phase are the sources, their decoding set corresponds to itself and the relay nodes have an empty decoding set.