FR2939584A1 - Network coding method for wireless mesh communication network, involves constructing output notification identifying differences between combinations of source packets, and transmitting modified result packet and output notification - Google Patents

Abstract

The method involves detecting whether input notification indicates that a used data packet is missing or is received in a modified manner such that linear combination of source packets forming the modified used packet is not conformed to a predetermined network coding scheme. A modified result packet is constructed in case of positive detection. An output notification identifying differences between the combinations of packets representing predetermined and modified result packets, respectively, is constructed. The modified result packet and the output notification are transmitted. Independent claims are also included for the following: (1) a computer program product comprising program code instructions for implementing a network coding method (2) a storage medium for storing a computer program comprising instructions for implementing a network coding method (3) a relay node comprised in a mesh network and participating in a network coding method (4) a destination node comprised in a mesh network and participating in a network coding method.

Description

Network encoding method in a mesh network, computer program product, storage means and corresponding communication nodes. FIELD OF THE INVENTION The field of the invention is that of communication networks. More specifically, the invention relates to a method for notifying potential changes with respect to a predetermined reference network coding scheme during an initialization phase of a communication network. 2. TECHNOLOGICAL BACKGROUND 2.1 Context and problem In the remainder of this document, the problem is described, in the context of a network coding in a wireless mesh network, which the inventors have faced. of the present patent application. The invention is of course not limited to this particular context of application, but is of interest in all cases of network coding adapted to the transmission, via relay nodes, of data packets from source nodes to destination nodes. , in a mesh communication network. The 60GHz wireless communication systems are particularly well suited for transmitting data streams with very high speed and limited range. At the time of the proliferation of high definition content, this type of wireless transmission and broadband is adequate, for example, as a connection between the various audio elements of a home theater (or home theater in English) requiring a bandwidth very important and this at the scale of a room. But, despite the high speeds implemented by these types of network, the multiplicity of sources however requires the increase of the available bandwidths in order to be able to transmit the contents of these different sources for example of the audio and the video always in a home cinema system. To overcome this network capacity problem, a new concept has emerged. This concept, called network coding (or network coding in English), allows a node not only to retransmit one of the packets it has received as input from one of the sources (or a relay node) but to create a resulting packet which is a combination of packets received as input.

The concept of network coding has been introduced by R. Ahlswede et al. and extensively described in a publication (R. Ahlswede, N.Cai, SY.R.L. Li, and RW Yeung, "Network Information Flow", IEEE Trans.Information Theory, Vol 46, pp. 1204-1216, July 2000 ). Very quickly, many articles have demonstrated the contribution of this new concept, especially in terms of bandwidth gain and in terms of network capacity utilization. In addition, it has been shown that in a multicast framework (or multicast in English) where each message sent by a source node is intended for all of the nodes of Dt (Dt being a set or all of the destination nodes of the network), the linear combinations in a sufficiently large body Fq give an optimal result (in terms of gain) as described in the document "Linear network coding," - S.-Y. R. Li, R. W. Yeung, and N. Cai, IEEE Trans. Information Theory, IT-49 (2): 371-381, Feb. In the remainder of the description, a network coding scheme is defined as follows: the relay nodes combine the input packets to generate a combined packet called the resulting packet; and for these relay nodes, the input packets to be combined are chosen from the set of packets received; and for each input packet to combine is associated a coefficient in the body Fq. The nodes and links participating in the network coding are considered useful, the others are considered redundant. The destination nodes then receive a plurality of packets (the resulting packets from the relay nodes) which are linear combinations of the source packets sent by the source nodes.

In one embodiment of the invention, the coding scheme used is an optimized coding scheme as described in US2005 / 0010675. This optimized coding makes it possible to reduce the number of nodes and useful links participating in the network coding scheme in order to reduce the number of operations and therefore the CPU time (for Central Processing Unit in English or central processing unit in French). This scheme is determined at the initialization of the system taking into account the topology of the system. The resulting packets transmitted by the relay nodes are therefore predetermined (more precisely their associated linear combination is predetermined).

However, because the wireless environment is very sensitive, transmission can be disrupted by many factors such as interference, attenuation, or masking caused by moving objects crossing the transmit field. These disturbances of the transmission channel can lead to loss of links and therefore temporary changes in the network topology. Since the predetermined network coding scheme is based on the initial topology of the network, a modification of this topology can lead to the inability of the destination nodes to decode the set of source packets because they are all interconnected in each other. received packets.

2.2 Prerequisites and limitations of the state of the art

2.2.1 The body Fe q = 2s

All the operations performed in the present invention is carried out in a finite body denoted Fq, consisting of q elements. In the context of the invention, the information sequences sent in the network are in a container called packet, which has a fixed size, s consecutive bits of a packet constituting a symbol on the body Fq with q = 2s. For example, a 1400-byte packet contains 1400 symbols if the body size is 28 or contains 700 symbols if the body size is 216. The addition, multiplication, and division operations are performed as follows: - Addition + is done bit by bit with the classic OR exclusive (or XOR) function

- The multiplication . is symbol to symbol. A symbol of s bits, b0 ... bs-1, is transformed into a polynomial: b0 + blX + .. bs-lXs-1. An irreducible polynomial on Fq is chosen. The multiplication between two symbols is then obtained by the classical multiplication between two polynomials modulo the irreducible polynomial. - The division is calculated from an Euclidean algorithm. Therefore, in the rest of the document, when one writes: Y = ai.Ar = 1 with A1 ... A1, 1 input packets and al, .., a1 coefficients in Fq, these operations take place in fact, on all the symbols of the packet: yk = ka; .ai i = l where yk and ak are respectively the k1th symbol of packets Y and A. The + and. take place in Fq. 2.2.2 Network coding 2.2.2.1 Introduction At a relay node, the linear combination associated with a resulting packet can be seen at two distinct levels, a local level and a global level. To explain these two levels, suppose that a network coding scheme is predetermined for all the nodes and situate us in a combining relay node. This relay node Ri must then combine a set of received packets M1, M2, ..., M1. The resulting packet is then equal to: yj = la / Mi with ai, belonging to the finite field Fq. = 1 This linear combination is the local representation of the coding, hereinafter called local combination. But the packets to combine received as input can be themselves also resulting packets from other relay nodes. Let us denote by X1, ... XM the source packets sent by the source nodes. It is then clear that each packet flowing in the network is a linear combination of source packets X1, ... XM. The aforementioned packet Yi can also be represented in the following way: yj = g'Xi. i = 1 This representation is the global representation of the coding. It takes into account that each received packet Mi is also a resulting packet combining source packets. It is called thereafter global combination. It is also clear that the coefficients g3 are composed of a combination of ai representing the different local combinations of the nodes upstream of the node receiving this packet in the disjoint path (the notion of disjoint path is more fully described later). It is then possible to write: g3 = 11a The products of a j are called group of coefficients.

Another representation of the resulting packet is to express it as a function of the source packets X; and coefficients ai: y; = 1 (11160X

This representation is subsequently called an explicit global combination.

Packets having non-zero coefficients in a linear combination (local, global or global explicit) are hereinafter called packets associated with this combination. In the context of the invention, each relay stores the global representation as well as the explicit global representation of each packet received as input in a matrix, respectively called relay matrix R = (rij) and explicit relay matrix Rd. Each line represents a packet. received, each column a source package. For the relay matrix R, each element of the matrix rij represents the coefficient of the source packet j of the global linear combination of the packet i. For the explicit relay matrix Rd, each element of the matrix corresponds to the explicit global representation of the coefficient associated with the source packet j contained in the packet i. The destination node Dj which receives m 'packets Y1, Y2, ..., Ym, also stores the global and global explicit representation of each useful packet received in a matrix, respectively called destination matrix D = (dij) and explicit destination matrix. dd. Each line represents a useful packet received and each column a source packet. For the destination matrix D, each element dij of the matrix represents the coefficient of the source packet j of the global linear combination of the packet Yi received. For the explicit destination matrix Dd, each element of the matrix corresponds to the explicit global representation of the coefficient associated with the source packet j contained in the packet Yi received. The destination node must then solve the linear system constituted by the set of global linear combinations: M Yi = X; 1 = 1 j = 1 ... m '. From a matrix point of view, Y = D.X with D the destination matrix, Y the column vector of the received packets and X the column vector of the source packets. Note that if the rank of D (in the linear algebra sense) is equal to the number of source nodes then all source packages are decodable. Otherwise, it is necessary to study more precisely the destination matrix to determine the number of decodable source packets as more fully described in relation to FIG. 10. It should be noted that the use of the explicit matrices is only useful in one embodiments of the invention (indirect notification mode), in the other mode these matrices are useless. 2.2.2.2 Construction of network coding In a mesh network consisting of M sources and N destination nodes, the preliminary step before building a network coding scheme is to ensure that there is a suitable network coding scheme to the topology of the network. This phase is based on graph theory concepts and in particular on a use of the Ford-Fulkerson algorithm more fully described in the document "Maximum flow through a Network" Canadian Journal of Mathematics, L. Ford R.Fulkerson, 8 , pp 399-404, 1956. The network is modeled by a graph G = (V, E), V being a finite set of K nodes (called vertices in English), V = {xl,. . . , xK}, and E V V x V a finite set of edges that connect the nodes. In the context of the invention, the set E is an ordered set of pairs (xi, xj), with xi and xj belonging to v, these pairs being called connected edges, arcs or links (called edges in English). A node of V models a communicating entity in the network (computers, speakers, amplifier) while an arc (xi, xj) indicates a sufficiently good radio communication (that is to say above a certain threshold of received radio power) between the two communicating entities xi and xj. Moreover, a path between a node xl and a node xp is called an ordered set of nodes {xl, x2, .., xp} such that (xi, xi + 1) E E for i = 1,. . . , (pùl). A path can also be written {(xl, x2), .., (xi, xi + 1), ..., (xp-1, xp)} with (xk, xk + 1) EE for k = 1 , .. , (pù1). We call k disjoint paths, k sets that do not have two consecutive nodes in common, either from a graphical point of view, two arcs in common. Let T be a set of k disjointed paths. We then define: (x, y) EE (a link of the graph), (x, y) AND if and only if there exists a path of T that contains (x, y), that is x EV (a vertex of graph), x AND if and only if there exists a path of T that contains x. We call two disjoint paths between a set S and a node d, two ordered sets of nodes {xl, x2, .., xp, d} and {yl, y2, .., yq, d} such that: - (xi , xi + 1) EE for i = 1, ..., (p1) and (xp, d) EE; - (yj, yj + 1) EE for j = 1, ..., (q1) and (yq, d) EE; - xl ES, yl ES and xl ≠ yl; `di = 1 .. (p -1),` dj = 1 .. (q -1), (xi, xi + 1) ~ (yj, yj + 1) and (xp, d) ~ (yq, d ); The definition is generalized very easily for a set of k disjoint paths. From these notions of graph theory, the state of the art gives the following condition: Let S be a set of M source nodes. If there are M disjoint paths between S and each destination node of Dt, then the M nodes are capable of simultaneously transmitting multicast to Dt if the relay nodes are allowed to linearly combine their incoming packets on a sufficiently large finite field. This result is a direct result of the article by R. Ahlswede and co. "Network information flow". The invention is placed in the context where this condition is verified at the start of the system. On the other hand, during the life of our system, this condition may no longer be respected because of temporary loss of links. In this case, no network coding scheme exists to allow the M source packets to be decoded. The destination nodes can no longer decode M source packages. Their decoding algorithm must then handle this case and in particular must allow to decode the maximum of source packets (strictly less than m).

In the state of the art, there are mainly two axes for the construction of a network coding scheme depending on whether or not the initial topology of the network is known. For a known topology, a first deterministic scheme for constructing coefficients based on a matrix approach (called a deterministic coding scheme) is given in the article by R. Koetter and M. Médard, "An Algebraic Approach to Network Coding" ACM Transactions on Networking, Vol 11, Nos, October 2003.

For an unknown topology, an on-the-fly construction of the randomly drawn coefficients (called the random coding scheme) is given in the article "Randomized distributed network coding" by Michelle Effros, Tracey Ho, David Karger, Ralf Koetter and Muriel Medard.

In the first approach, the coefficients are calculated once and for all at the initialization of the system and thus valid for the lifetime of the system. The destination nodes and the relay nodes have inexpensive operations to perform because they know in advance the packets they will receive and transmit. In the second approach, the relay nodes randomly pull the coefficients of the resulting packet to be generated at each new packet to be transmitted, which requires additional processing time. In addition, since the coefficients are different for each new packet generated, the relay nodes must transmit them in addition to the packet so that the destination nodes are aware of the new linear combinations and can, as far as possible, decode the source packets. But this signaling is even larger than the body is large (each coefficient will be encoded on a larger number of bits), and it is as much lost bandwidth. But in this second approach, the larger the body, the greater the chances of successful decoding of source packages. Hence the contradiction. On the other hand, in the first approach, the decoding is sure to succeed because the construction performed at initialization ensures correct decoding. The first approach is therefore more efficient than the second in terms of gain in bandwidth and computing time. The second approach has some advantages in terms of robustness to link loss. Indeed, as this method generates a resulting package with a different linear combination with each send, it does not require knowledge of the network and the loss of link is transparent to it (although not guaranteeing decoding). It indicates a zero coefficient for the packet that was to arrive by this link. In the context of the invention, there is a way to recover the topology of the network and thus to build a network coding scheme with the first method. However, the first approach is strongly affected by link loss. Indeed, this link loss modifies the linear combination of the relay node initially fixed and only the relay node observing this loss has knowledge thereof. The other nodes and especially the destination nodes are not aware of this modification of the topology and believe that everything is happening as usual. Decoding can then be catastrophic and affect all the source packets to be decoded since they are all linked by linear combinations (unlike a routing or only a source packet would be reached). Considering that the advantages of the first approach are essential in high-speed applications as targeted by the invention, the technical problem solved by the present invention is to construct a predetermined network coding and calculated from the initial topology, robust to broken links. . In practice, the method of constructing the predetermined coding scheme (i.e. the first approach) commonly used is that which limits the number of nodes and links participating in the network coding using only the disjoint paths of the condition of existence of network encoding (only nodes and links of disjoint paths are useful for building the network encoding scheme). This concept is described in particular in the patent document US2005 / 0010675 of the company Microsoft Corporation. On the basis of this observation, we call a node (or link) that belongs to the set of disjoint paths and, we call a node (resp link) redundant, a node (resp. ) that does not belong to the set of disjoint paths. This optimized version thus makes it possible to limit the number of operations. But in a wireless LAN, it is likely that these redundant links are numerous, because every time a node sends a packet, all other nodes in the network are likely to receive this packet. These redundant packets will not be taken into account by the nodes (i.e. they are not stored).

This optimized version of the network construction is used in the context of our invention. It severely limits operations because only the nodes and links considered useful are used. For example, a relay node no longer needs to combine the plurality of packets it receives but only a subset of packets. This is also true for the recipient node which only needs a subset of the packets received to decode the source packets. Therefore the invention will initially build a network coding scheme that takes into account only the nodes and useful links. The relay nodes store and therefore combine only the useful packets with which a coefficient belonging to the body Fq is associated. The generated packets called the resulting packets are predetermined. In the same way, the destination node constructs a destination matrix D, composed only of the useful packets (the others not being stored). In order to decode the source packets, the node calculates the inverse of the destination matrix to obtain the following linear system: X = D-1Y with X the column vector of the source packets and Y the column vector of the packets received. Note: The Gaussian Pivot method is also used to calculate the rank of a matrix. For this, we perform a triangulation of the matrix. Then, the rank of the matrix is easily obtained by counting the number of non-zero rows of the matrix. 2.2.3 Limits of the state of the art In the state of the art, the management of packet losses in a network coding scheme is mainly managed by the random coding scheme as previously described where the relay nodes systematically perform expensive operations both in terms of computation time to combine the packets with or without packet loss as the bandwidth used to transmit the value of the coefficients. These disadvantages minimize the benefits sought in terms of bandwidth gain with network coding. In the context of deterministic coding scheme presented in R. Koetter's article, M. Medard, "An Algebraic Approach to Network Coding" ACM Transactions on Networking, Vol 11, Nos, October 2003 and in the article by PA Chou , Y. Wu, and K. Jain. "Practical network coding," In Proc. 41stAllerton Conf. Communication, Control and Computing, Monticello, IL, Oct. 2003 ", the state of the art is looking for robust coding schemes as soon as they are built, trying to ensure that destination matrices are always able to to decode all the source packets sent by the source nodes for a predefined set of broken links A disadvantage of this prior art technique lies in the fact that it is not possible to use a coding scheme This is because, with this known technique, it is mandatory to use all the received packets, which imposes a large and especially systematic calculation time on the relays, which is also detrimental in high speed applications in which There is a large flow of data to be processed 3. OBJECTIVES OF THE INVENTION The invention, in at least one embodiment, has the particular objective of overcoming the various disadvantages of the state of the art. More specifically, in at least one embodiment of the invention, one objective is to provide a technique for performing a robust network coding scheme for link loss and also for the destination node (also referred to as the destination node). to adapt his decoding scheme. A complementary objective of at least one embodiment of the invention is to provide such a technique for optimizing CPU resources. Another objective of at least one embodiment of the invention is to provide such a technique which is simple to implement for a lower cost. SUMMARY OF THE INVENTION In a particular embodiment of the invention, there is provided a method of network coding in a mesh network comprising at least one relay node which, according to a predetermined network coding scheme, expects receiving as input a plurality of payload packets and outputting a predetermined resultant packet obtained according to a predetermined linear combination of said plurality of payload packets, each of the payload data packets being formed according to said predetermined network coding scheme by a linear combination of one or more source packets, the source packets being transmitted by source nodes. The method is implemented by at least one relay node and comprises the steps of: - detecting whether at least one input notification indicates that a useful packet is missing or has been received modified so that the linear combination of the source packet or packets forming said modified useful packet does not conform to the predetermined network coding scheme; - in the case of a positive detection, construct a modified resulting packet from the received useful packets; Constructing an output notification identifying the differences between the linear combination of source packets representative of the predetermined resulting packet and the linear combination of source packets representative of the modified resulting packet; and - transmit the resulting modified packet and the output notification. Thus, in this particular embodiment, the invention is based on a completely new and inventive approach to a network coding method of notifying potential changes with respect to a predetermined reference network coding scheme established during a phase. initialization of the communication network for example. For this, the relay node impacted by a link loss sends a resulting packet different from the predetermined resulting packet due to a missing useful packet, but notifies this difference through an output notification. This exit notification will be an input notification for an intermediate relay node or a destination node. The input notification allows the intermediate relay node to build in turn a modified resulting packet. In the end, the input notification allows a destination node to take into account the new combination of the packet in order to modify its predetermined decoding scheme. The modified coding scheme makes it possible to recover at least the source packets that are not missing in the received source packet combination and thus avoid corrupting these packets by an inappropriate decoding scheme. As a result, the present invention provides an alternative and temporary coding scheme for the detection of a loss of link which makes it possible to limit the impact of variations in the topology of the network.

Thus, in a simple and effective manner, a network coding scheme tolerant to loss of links (unwanted modification of the predetermined network coding scheme). Advantageously, the input notification represents an event internal to the relay node when it indicates that a useful packet is missing; or information received with a useful packet when it indicates that said useful packet has been received modified.

A particular embodiment of the invention provides that each input or output notification comprises M fields, with M the number of source nodes, each of the M fields being associated with one of the M source packets generated by the M nodes. sources, each of the M fields comprising: a first subfield indicating whether a coefficient, associated with the source packet associated with said field, has been modified or not, in said linear combination of source packets representative of said modified resulting packet, with respect to said linear combination of source packets representative of said predetermined resulting packet; a second subfield providing information relating to a value of the modified coefficient to which the first subfield relates, if the first subfield indicates a modification. Thus, each source packet is associated with a notification to inform whether a coefficient of the linear combination of packets has been modified or not. This notification then allows a destination node receiving the notification to modify its source packet decoding scheme and a relay node to modify its output notification and its resulting modified packet. Advantageously, the second subfield is present only if the first subfield indicates a modification.

Thus, the number of data to be transmitted over the network is reduced. The bandwidth is thus optimized. Also, the information included in the second subfield explicitly includes the value of the modified coefficient. Thus, the number of operations inherent in calculating the coefficient for the node receiving a notification is decreased. The size of the memory is also reduced. Advantageously, the information included in the second subfield comprises at least one indication enabling a node processing said entry or exit notification to recover the value of the modified coefficient. Thus, the invention makes it possible to choose the mode of notification minimizing the amount of information to be transmitted over the network and thus bandwidth gain.

In a particular embodiment of the invention, there is also provided a method of network coding in a mesh network comprising at least one destination node which, according to a predetermined network coding scheme, expects to receive as input a plurality of useful data packets and must output a plurality of decoded source packets each obtained in a predetermined linear combination of said plurality of payload packets, each of the useful data packets being formed, according to said predetermined network coding scheme, by a linear combination of one or more source packets, the source packets being transmitted by source nodes. The method is implemented by at least one destination node and includes the steps of: - detecting whether at least one input notification indicates that a useful packet is missing or has been received modified so that the linear combination of the source packet or packets forming said modified useful packet does not conform to the predetermined network coding scheme; in the case of a positive detection, obtain a modified network coding scheme, according to the linear combinations of source packets forming the received useful packets; and decoding the received payload packets based on said modified network coding scheme to obtain at least one decoded source packet. Thus, in this particular embodiment, the invention is based on an entirely new and inventive approach to a network coding method of notifying the destination node to modify its predetermined decoding scheme in order to take into account the packets received. modified and thus decode a maximum of source packages. According to an advantageous characteristic of the invention, the step of obtaining a modified network coding scheme comprises the following steps: modifying a reference destination matrix to obtain a modified destination matrix, according to said at least one notification detected input; detecting whether the rank of said modified destination matrix is equal to the total number of source packets to be decoded; and if the rank of said modified destination matrix is equal to the total number of source packets to be decoded, use said modified destination matrix to generate all decoded source packets in said step of decoding. Thus, all source packets are decodable by the destination node.

According to another advantageous feature of the invention, if the rank of said modified destination matrix is not equal to the total number of source packets to be decoded, the step of obtaining a modified network coding scheme comprises steps of : - determining a sub-matrix of said modified destination matrix, whose rank is different from zero; using said sub-matrix to generate at least one decoded source packet, in said step of decoding. Thus, although only part of the source packets is decodable by the destination node, the decoding of these latter by the destination node remains possible.

According to one particular interesting feature of the present invention, the input notification represents an event internal to the relay node when it indicates that a useful packet is missing; or information received with a useful packet when it indicates that said useful packet has been received modified. Advantageously, each input notification comprises M fields, with M the number of source nodes, each of the M fields being associated with one of the M source packets generated by the M source nodes, each of the M fields comprising: a first sub field indicating whether a coefficient of the source packet associated with said field has been modified or not, in said linear combination of source packets representative of an expected useful packet; a second subfield providing information relating to a value of the modified coefficient to which the first subfield relates, if the first subfield indicates a modification. Thus, each source packet is associated with a notification to inform whether a coefficient of the linear combination of source packages has been modified or not.

This notification allows the destination node receiving said notification to modify its predetermined decoding scheme of the source packets.

Advantageously, the second subfield is present only if the first subfield indicates a change. Thus, the number of data to be transmitted over the network is reduced. The bandwidth is thus optimized.

Also, the information included in the second subfield explicitly includes the value of the modified coefficient. Thus, the number of operations inherent in calculating the coefficient for the destination node receiving the notification is decreased. The size of the memory is also reduced.

According to another characteristic, the information included in the second subfield comprises at least one indication enabling a node receiving said input notification to retrieve the value of the modified coefficient. The invention thus makes it possible to choose the notification mode minimizing the amount of information to be transmitted on the network, which makes it possible to obtain a gain in bandwidth. In another embodiment, the invention relates to a computer program product downloadable from a communication network and / or recorded on a computer readable medium and / or executable by a processor. This computer program product includes program code instructions for carrying out the aforesaid method (in any one of its various embodiments), when said program is run on a computer. In another embodiment, the invention relates to a computer readable storage means, possibly totally or partially removable, storing a computer program comprising a set of instructions executable by a computer to implement the aforementioned method (in any of its different embodiments). The invention also relates to a relay node included in a mesh network and participating in a network coding method in said mesh network, said relay node, according to a predetermined network coding scheme, expecting to receive as input a plurality of packets of useful data and having to output a predetermined resultant packet obtained in a predetermined linear combination of said plurality of payload packets, each of the payload packets being formed, according to said predetermined network coding scheme, by a linear combination of a or more source packets, the source packets being issued by source nodes. Said relay node comprises: detection means, for detecting whether at least one input notification indicates that a useful packet is missing or has been received modified so that the linear combination of the source packet (s) forming said modified useful packet does not conform to the predetermined network coding scheme; first building means, activated in the case of a positive detection, making it possible to construct a modified result packet from the received useful packets; second construction means for constructing an output notification identifying the differences between the linear combination of source packets representative of the predetermined resulting packet and the linear combination of source packets representative of the modified resulting packet; transmission means for transmitting the modified resulting packet and the output notification. The invention also relates to a destination node included in a mesh network and participating in a network coding method in said mesh network, said destination node, according to a predetermined network coding scheme, expecting to receive as input a plurality of packets of useful data and having to output a plurality of decoded source packets each obtained in a predetermined linear combination of said plurality of payload packets, each of the data packets being formed, according to said predetermined network coding scheme, by a linear combination of one or more source packets, the source packets being transmitted by source nodes. The destination node includes: - detection means for detecting whether at least one input notification indicates that a useful packet is missing or has been received modified so that the linear combination of the source packet (s) forming said modified useful packet does not conform to the predetermined network coding scheme; means of obtaining, activated in the case of a positive detection, making it possible to obtain a modified network coding scheme, as a function of the linear combinations of source packets forming the useful packets received; decoding means, for decoding the received useful data packets according to said modified network coding scheme, to obtain at least one decoded source packet. 5. LIST OF FIGURES Other features and advantages of the invention will appear more clearly on reading the following description of a preferred embodiment, given by way of a simple illustrative and nonlimiting example, and the appended drawings, among which: - Figure 1 illustrates schematically, according to a particular embodiment of the invention, an example of wireless mesh network some nodes implement a network encoding; FIG. 2 schematically illustrates the architecture of a communication module according to a particular embodiment of the invention; FIG. 3a schematically illustrates the algorithm for constructing and broadcasting, by a central node, the predetermined coding scheme according to one particular embodiment of the invention; FIG. 3b schematically illustrates the algorithm for reception, by a relay node, of the coefficients of the predetermined coding according to a particular embodiment of the invention; FIG. 3c schematically illustrates the algorithm for reception, by a destination node, of the coefficients of the predetermined coding according to a particular embodiment of the invention; FIG. 4 schematically illustrates the structure of a notification notifying a modification of the predetermined coding scheme according to a particular embodiment of the invention; FIG. 5 diagrammatically illustrates the main algorithm implemented in a relay node according to a particular embodiment of the invention; FIG. 6 schematically illustrates the detailed algorithm of the construction of a resulting packet according to a particular embodiment of the invention; FIG. 7 diagrammatically illustrates the main algorithm implemented in a destination node according to a particular embodiment of the invention; FIG. 8 diagrammatically illustrates the algorithm for recovering the coefficients of the received packets modified by a relay node or a destination node, according to a particular embodiment of the invention; FIG. 9 diagrammatically illustrates the detailed decoding algorithm implemented by a destination node according to a particular embodiment of the invention; - Figure 10 schematically illustrates the search algorithm of a decodable sub-matrix according to a particular embodiment of the invention. DETAILED DESCRIPTION With reference to FIG. 1, a preferred embodiment of the invention is presented in the case of a Mesh wireless communication system composed of several nodes applying the network coding as shown in FIG. FIG. 1. The access protocol to the medium is a TDMA time access (for Time Division Multiple Access or in French) where during a fixed period, called superframe, each node of the network has unique access to the medium during a fixed speaking time (or slot in English). This network is composed of three source nodes (represented in black) named respectively NS1, NS2 and NS3 and six nodes being both relay nodes and destination nodes (NR / D1, .., NR / D6). Each communication link is represented by an arrow indicating the direction of the transmission. The full-line communications links, such as between NS1 and NR / D1, represent packets composed solely of the source packet derived from NS1. The small dotted communication links such as between NS2 and NRID4 represent packets composed only of the source packet from NS2 and the large dotted communications links such as between NS3 and NRID2 represent packets containing only the source packet. NS3. The other communications links (e.g. between NR / D1 and NRID3) represent packets that have undergone network coding and therefore contain a linear combination of at least two source packets. The other unrepresented communications links are either victims of permanent masking, or these links are not usable because of the TDMA protocol and the successive speech times given to each node, the source nodes emitting first in the order of their numbering. and then all the relay nodes also in the order of their numbering. Each packet is in the form described in connection with FIG. 4. The packet consists of two parts: A part 410 in which the information emitted by the source nodes is contained. This part is a linear combination of the source packets sent by the source nodes, this linear combination being predetermined at the initialization of the system as more fully described below in relation to FIG. 3. A so-called notification part 405 even composed of M under parts (M being the number of sources) and each of these M subparts is composed of two fields. The first field (415, ..., 425) is an indicator as to whether the predefined coefficient of the predetermined linear combination associated with the source packet has been modified or not. This field is always present and could be composed of a bit, the value 0 corresponding to no modification on the coefficient and the value 1 to modifications on the coefficient of the associated source package. Depending on the value of the indicator field (typically 1 in the previous example), an Info_Coeff field (420, ..., 430) is present or not in the notification. This field gives sufficient information to the node receiving this notified packet to find either directly or indirectly the new value of the coefficient (s) associated with the source packet (s) represented by the () s) field (s) of the notification. The choice between these two modes (direct or indirect) is more fully described later. This field can be either directly the new value of the coefficient (direct mode), or the indication of the coefficients (of the explicit global form) not taken into account by the preceding node to calculate the received coefficient (indirect mode). One bit is allocated for each group of coefficients (of the explicit global form), the value 0 corresponding to a non-use and the value 1 to the use of this group of coefficients (of the explicit global form) in the calculation of the combination received. For example, the combination g5 = a5.a1 + a5.a3 + a5.a4 is sent instead of the predetermined combination g5 = a5 .a, + a5.a2 + a5.a3 + a5 .a4. The associated Info_Coeff field then takes the value 1011. A coefficient group represents the result of the multiplication of several coefficients (in the previous example: a5 .a,, a5 .a2, a5.a3 and a5.a4). The predetermined global coefficient is therefore a sum of coefficients group which represents the explicit global combination. 6.1 Device of the invention The device according to a particular embodiment of the invention is illustrated in FIG. 2 by the block 210. An application 230 and a radio frequency transmitter / receiver 240 correspond to the environment of the block 210. Man / Machine interface 212 allows the user to give the graph representing the network in order to compute the reference network coding scheme by means of a Ford Fulkerson algorithm. This network coding scheme will then be shared between all the nodes of the network. A given memory 216 makes it possible to store the different copies of the data transmitted by the sources and the different relay nodes. This data received by the Radio Frequency Transmitter / Receiver 240 passes through a radio packet receiver 215. This memory also makes it possible to store the matrices of the various nodes involved in the network coding process as described below. The device block 210 also comprises a module 217 for managing network coding packets. This module segmented into four sub-modules 218, 219, 220 and 221 manages the network coding packets both at the relay nodes and the destination nodes. The sub-module 218 is a notification analyzer that makes it possible to know whether the received packet corresponds to the expected packet or if coefficients related to the network coding have been modified.

The sub-module 219 allows the update of the network coding coefficients to be used either for the retransmission or for the decoding instead of the reference coefficients stored in the data memory 216. The update of these coefficients is carried out after the analysis of a notification corresponding to a received packet. The coded packet generation sub-module 220 is used by the relay nodes. It makes it possible to generate the resulting package that must be sent. This resulting packet is a function (linear combination) of some received packets and their associated coefficients. Once produced, the resulting packet is stored in the data memory 216 to be transmitted by the radio packet transmitter 214. The sub-module 221 is used during the decoding phase. It makes it possible to construct a decoding matrix according to the received packets and their associated coefficients stored in the data memory 216 making it possible to extract from this matrix the different source packets. Once decoded, the source packets are transmitted to the application 230 thanks to the CPU 211. 6.2 Initialization of the system FIGS. 3a, 3b and 3c describe algorithms, implemented in a central node, a relay node and a destination node respectively , of construction of a predetermined network coding scheme, at system startup. The central node, which is one of the nodes present in FIG. 1 (for example one of the source nodes), has the function of constructing the predetermined network coding scheme and of distributing it to all the other nodes (as described in FIG. Figure 3a). On receipt of this network coding scheme by the other nodes of the network, the relay nodes store the latter (as described in FIG. 3b) and the destination nodes as well (as described in FIG. 3c). We now present, in relation with Figure 3a, the algorithm implemented in the central node. A first step 305 corresponds to the initialization of the system. Then in a step 310, the central node retrieves the topology of the network, that is to say, it retrieves the quality of the communications between the nodes, the list of source nodes and the list of destination nodes. This topology can be known from the nodes of the network by construction if the network is fixed and if the number of nodes is constant. It is also certain that this information can be introduced by the user after the installation of a system.

Then in a step 315 is built a network graph G from the topology of the network. If the quality of a transmission between two entities is considered good, for example if the signal-to-noise ratio (or Signal Noise Ratio) is greater than a given threshold (depending on the type of radio used), the transmission is considered valid. Otherwise, it is considered invalid. The network graph G is then constructed as follows: each communicating entity is modeled by a node of the graph and each communication considered valid between two entities is modeled by a link between two nodes (and therefore each communication considered invalid is not not modeled in the graph).

Then a step 320 determines the predetermined coding scheme. The method is explained in US2005 / 0010675 of Microsoft Corporation and repeated in many state-of-the-art articles. This method described in US2005 / 0010675 is cited as a teaching element and not as a fundamental step of the present invention. The steps implemented in this method are the following: a step of searching the disjoint paths Uj between the set S of the source nodes and each destination node Dj; a step of constructing a subgraph G 'of G containing only vertices and links belonging to at least one set Uj; a step of determining the coefficients of the linear combinations in such a way that the set of destination matrices is of rank equal to the number of sources. The size of the body q is therefore implicitly determined as well as the maximum number of groups of coefficients contained in the linear combination of the resulting packet of each relay node p.

A next step 325 makes it possible to choose the notification mode used in the Info_Coeff field (420, ..., 430) of the notification (FIG. 4). Two modes are conceivable: - a direct mode in which the coefficients are explicitly described in the notification; and an indirect mode in which the notification indicates the impacted coefficients, the relay nodes must then determine the new value of the coefficients. The chosen mode is the one that will be the least expensive in terms of bandwidth. The choice between the two modes is then done by comparing the number of bits necessary to code the coefficients of the network coding (that is to say the size of the body necessary for the decoding matrices to be invertible) with the maximum number of packets containing the same source packet that a relay node will have to combine (the maximum number of aj coefficients, composing the explicit global combination of a given source packet, contained in the linear combination of the resulting packet of each relay node), and taking the one that will minimize the number of bits needed for the notification. This test can be written in the following way: Mxq> _Mxp with M the number of sources of the network, q the size of the body and p the maximum number of group of coefficients, composing the explicit global combination of a given source package, contained in the linear combination of the resulting packet of each relay node. Although the selection criterion for the implementation of the present invention seeks to minimize the size of the transmitted notification, the choice of the notification mode may also be influenced by the context of the system. Indeed, the direct mode reduces the number of operations inherent to the calculation of the coefficient for the node receiving the notification (the new coefficient being already calculated). In addition, the indirect mode requires that each node of the network has knowledge of its relay matrix (or destination) in its explicit global form in order to find the new coefficient which requires the use of a larger memory in comparison with the direct mode.

Then a step 330 implements a diffusion protocol of the coefficients and the notification mode. Several diffusion techniques are known to those skilled in the art for broadcasting information in a mesh network, for example flood data type in a mesh protocol (or flooding protocol in English).

At the end of this diffusion step 330, each relay node receives in a step 340 the coefficients of the detailed linear combinations of all the nodes of the network and the notification mode that will be used. FIG. 3b diagrammatically illustrates the algorithm implemented in a relay node on reception of the predetermined network coding scheme broadcast by the central node (FIG. 3a). Each relay node keeps in memory the combination that it will have to perform thereafter, and builds in a step 345 a relay matrix and an explicit relay matrix (the explicit relay matrix is constructed only if the notification mode is the indirect mode ). The explicit relay and relay matrices respectively contain the global combinations and the explicit global combinations of the resulting packets of the source nodes or relay nodes that had access to the medium during the superframe before the access date to the medium of the relay node considered. The other packets, ie those sent by nodes which had access to the medium after the relay node, are not taken into account because they are not usable (because of the constraint that the combinations perform only with data sent during the current superframe). FIG. 3c schematically illustrates the algorithm implemented in a destination node on reception of the predetermined network coding scheme broadcast by the central node (FIG. 3a). At the end of the diffusion step 330 (FIG. 3a), each destination node receives in a step 350 coefficients of the detailed linear combinations of all the nodes of the network. Each destination node then constructs, in a step 355, a destination matrix D and an explicit destination matrix (the explicit destination matrix is constructed only if the notification mode is the indirect mode).

In a step 360, from the destination matrix and the Gauss Jordan algorithm, the inverse of the destination matrix D is calculated (D is by definition invertible). The destination node then stores the inverse of the destination matrix D, a matrix called a decoding matrix that the destination node will use directly on receiving the packets. The different matrices are more fully described later. After this initialization phase, each source node starts sending the source packets on the network, each relay node performs its local combination and each destination node decodes the source packets from its decoding matrix. 6.3 Relay Node Figure 5 describes an algorithm for receiving a source packet by a relay node. In a first step 500, the relay node is waiting to receive a packet. In a next step 505, on receiving a packet, the relay node stores the packet and its associated notification in a step 520. Once the packet is stored, the relay node checks in a step 525 whether the received packet was the last one. before his own speaking time.

If so, the node builds in a step 530 its resulting packet to emit. This step 530 is described more fully later in connection with FIG. 6. If this is not the case, the relay node resumes waiting to receive a new packet (step 500).

In the case where a packet is expected but not received, an internal event allows the relay node to perform a step 510 in place of step 505. The detection of a loss of link is made by making a comparison between an RSSI measure and a threshold. If the measured RSSI is below the threshold then the packet is considered not received. In a step 515, the relay node then creates an ALL_MISSING notification associated with the expected packet but not received. The ALL_MISSING notification means that all the coefficients of the expected but not received packet are set to 0. The packet is then saved with its notification in step 520. In the case of expected but not received packets, only the notification is saved. It should be noted that the current relay matrix is reset with the reference relay matrix after each transmit cycle of all the nodes. FIG. 6 diagrammatically illustrates, according to a particular embodiment of the invention, an algorithm for constructing, by a relay node, a resulting packet to be transmitted (step 530 of FIG. 5). The construction of the resulting packet to be issued is broken down into three parts: a first part corresponding to the analysis of the modification notifications; a second part corresponding to the modifications of the current relay matrix in the event of notification of modification; and a third part corresponding to the construction of the packet to be transmitted. The first part corresponding to the analysis of the notifications starts in a step 600 by the reset, by the relay node, of its packet index to combine Combine packet_Index and of its local notification Notif. Then, in a step 605, the node tests whether the index of the packet to be combined is less than the number of packets that it must combine to create its resulting packet to send. If the test of step 605 is positive, the system checks in step 610 the presence of coefficient changes in the notification by analysis of the different fields. Indicator (415, ..., 425) of the packet notification Combine packet_Index . If no modification is present (all the fields are at 0), in a step 625, the system increments the variable Combine Packet_Index by one.

Otherwise, the local notification is set to 1 in a step 615 and the system updates in a step 620 the coefficients of the bundle packet_Imp_ packet received notified. This step 620 is more fully described later in connection with FIG. 8. If the test of step 605 of the comparison between Combine Pack_Index and the number of packets to be combined is negative, it means that all packets to be combined have been tested. The system therefore proceeds to the calculation of the resulting packet to be transmitted. The first calculation step begins in a step 630 of checking the value of the local notification Notif. If the test of step 630 is negative, the system builds in a step 645 the predetermined packet from the predetermined linear combination. Otherwise, in a step 635, the system builds the packet based on the new coefficients determined in step 620 (and more fully described later in connection with FIG. 8) and constructs the associated notification (output notification) in a next step 640. The construction of the notification in step 640 depends on the mode chosen for it at the initialization of the system. In both modes, the Indicator field associated with a source packet is set when the coefficient of the latter is changed in comparison with the coefficient of the linear reference combination. In this case, for the direct mode, the value of the new coefficient is put in the Info_Coeff field while for the indirect mode, the Info_Coeff field is filled by putting the value 1 to the groups of coefficients (of the global combination explained) used and the value 0 to the groups of coefficients (of the explicit global combination) not used for the calculation of the coefficient associated with the source package. In practice, the relay node reuses notifications of received packets. If the Indicator field for the source packet j of the packet i is 1, the relay node copies the information in the Info_Coeff field for the number of groups of coefficients associated with the source packet j otherwise the relay node fills the Info_Coeff field with 1 for the number of groups of coefficients. For example, a relay node receives two packets: M1 and M2. In the example, the notification has a size of 5 bits, defined at initialization as the maximum number of groups of coefficients contained in the linear combination of the resulting packet of each relay node, p, which in our case corresponds to the node relay creating the Mrésultant package. For messages whose coefficient group number is less than p, the least significant bits not assigned to a group of coefficients are set to zero. The global combinations explained are: - Mt = (a5.a1 + a5.a3) Xl + (a5.a2 + a5.a4) X2 - M2 = (a11.a7 + "11'0 (8) X1 + (" 11.0 (9 + a11 alo) X2 With X1 and X2 the two source packets And the notifications are respectively: - MlX1: 1 - 1000 - M1X2: 1 - 1000 - M2X1: 0 - M2X2: 0 The two least significant bits of M1X1 and M1X2 are set to zero because they do not correspond to any group of coefficients The relay node would send the following predetermined combination: Mrsul as predetermined = a12 ((a5 .a1 + a5 .a3 + a11.a7 + a11.a8) X1 + (a5 .a2 + a5 .a4 + a11.a9 + a11 .a10) X2) It will send: Mresultant = "12 ((" 5. "1 + a11.a7 +" 11.0 (8) X1 + (a5 a2 + a11.a9 + "11'0 (10) X2) and the associated notification: MresultantX1: 1 - 1011 MrsultantX2: 1 - 1011 Note that for Info_Coeff information to be valid for all nodes, it is necessary that the order of the groups of coefficients (of the explicit global form) is known by all the nodes of Figure 8 schematically illustrates, according to a particular embodiment of the invention, an algorithm for recovering the coefficients of the packet (step 620 of FIG. This algorithm updates the value of the coefficients from the different received notifications, thus obtaining modified coefficients. The algorithm begins by resetting a local Source Packet_Index variable in a step 800. In a next step 805, this Source Pack_Cluster variable is then compared to the value of the source number in the network. This test makes it possible to know if all the notification fields associated with the source package, for a given given packet, have been analyzed. If the test of step 805 is negative, it means that the analysis of the resulting packet is complete. The system then proceeds to a step 835 notifying the end of the algorithm. If the test of step 805 is positive, the system analyzes in a step 810 the field 415 Indicator (FIG. 4) of the notification associated with the source packet Source_package_Index. Depending on the value present in the indicator field 415, two cases arise: - either there is no modification of coefficients associated with the global combination explained (that is to say that the field is at 0); or at least one of the coefficients associated with the explicit global combination has been modified (i.e. the field is 1). In the first case, the system increments in a step 830 the variable Source Pack_Index and then returns to the test of step 805. In the second case, the system analyzes the field 420 Info_Coeff (FIG. 4) of the notification associated with the source package Source Pack_Index having been modified, that is to say that the system sets, in a step 820, the coefficients of the explicit global combination not present (marked with a 0 in the Info_Coeff field). For example, the Info_Coeff field takes the value 1011 and the predetermined explicit global combination is: g5 = a5.a, + a5.a2 + a5.a3 + a5.a4.

The node receiving the packet deduces that the predetermined overall coefficient g5 = a5.a1 + a5.a2 + a5.a3 + a5.a4 becomes g5 = a5.a1 + a5.a3 + a5.a4. Then, the system calculates the new coefficient and stores its value in the relay matrix (or the destination matrix if it is placed in a destination node) temporary (step 825). Then, in a step 830, the system increments the variable Source Pack_Index and then returns to the test of step 805. FIG. 8 describes the coefficient recovery algorithm more particularly for the indirect notification mode. For the direct notification mode, the steps 820 and 825 are replaced by a single step of reading the value of the new coefficient contained in the Info_Coeff field of the notification and by a step of storing this value in the matrix (relay or destination). 6.4 Destination node Figure 7 describes, according to a particular embodiment of the invention, an algorithm for receiving a source packet by a destination node. This algorithm is identical to that of the relay node of FIG. 5 except for steps 725 and 730. Step 725 verifies that the packet is the last to receive for the current superframe and step 730 is the decoding step plus amply described in connection with FIG. 9.

It should be noted that the current destination matrix is reset with the reference destination matrix after each decoding. FIG. 9 diagrammatically illustrates, according to a particular embodiment of the invention, an algorithm for decoding a packet received by a destination node (step 730 of FIG. 7). This algorithm is broken down into four successive parts: - a first part corresponding to the analysis of the notifications; a second part corresponding to the modifications of the current destination matrix in the event of notification of modification; a third part corresponding to the search for an invertible sub-matrix of maximum rank; and a fourth part corresponding to the decoding of the source packets. The analysis of the notifications begins in a step 900 by resetting, by the destination node, its current packet index Pack_Index and its local Notif notification.

Then, in a step 905, the destination node tests whether the index of the current packet is smaller than the number of packets useful for decoding. If the test of step 905 is positive, the system checks the different fields Indicator (420, ..., 430) of the notification of each packet Pack_Index. If no notification is present (all the fields are at 0), the system increments, in a step 925, by one unit the variable Pack_label. Otherwise, the local notification is set to 1 in a step 915 and the system updates the coefficients of the pack_pack_pack received packet notified in a step 920 (more fully described later in connection with FIG. 8). If the test of step 905 of the comparison between Pack_Index and the maximum number of packets used for the decoding is negative, it means that all the packets useful for the decoding have been tested. The system then checks in a step 930 whether the current destination matrix has undergone changes with respect to the reference destination matrix. In the case where no current packet has been modified, the system uses the inverse of the predetermined matrix for decoding in order to decode all the source packets in a step 970. Then, in a step 985, the system indicates that all source packages are available. The system then completes the decoding algorithm in a step 990.

If the test of step 930 is positive, i.e. current packets have been modified, the system performs, in a step 935, the calculation of the rank of the current destination matrix by the pivot method of Gauss. Then, in a step 940, the rank of the matrix is compared to the total number of sources to be decoded.

If the number of sources to be decoded is equal to the rank of the matrix, it means that the matrix makes it possible to decode all the source packets. To decode the source packets, the system calculates the inverse of the new matrix in a step 975, for example by the Gauss-Jordan method. The system then uses it, as well as the received packets, to decode all source packets in a step 980. Then step 985 is executed so that the system indicates that all source packets are available. The algorithm then ends with step 990. On the other hand, if the rank of the matrix is different from the number of source packets to be decoded (negative test in step 940), this means that the matrix does not make it possible to decode all source packages. The system should therefore find, if possible, a sub-matrix to decode a lower number of source packages. This sub-matrix search is performed in a step 945 (more fully described later in connection with FIG. 10).

Then a step 950 makes it possible to test how many source packets can be decoded by the system. If there are no decodable source packets (that is, if the rank of the matrix is zero), it means that no source packet can be decoded. decoding in a step 965. Then step 985 is executed so that the system notifies this failure to the organ performing the link with the application, for example the CPU. The algorithm then ends again with step 990. If the test in step 950 is negative, then the system calculates the inverse of the sub-matrix in step 955. In a step 960, the system then uses with the received packets associated with this sub-matrix to decode the source packets contained in the received packets. Next, the system marks the decoded source packets as being available in step 985, and then terminates the algorithm by step 990.

FIG. 10 diagrammatically illustrates, according to a particular embodiment of the invention, an algorithm for selecting the destination sub-matrix for decoding the most possible source packets (step 945 of FIG. 9). The output parameters are therefore the destination sub-matrix (if it exists) and the number of decodable source packets (rank of the sub-matrix). The system enters the algorithm if the current matrix does not allow to decode all the source packets; the algorithm tests combinations by first removing a number of sources until a decodable submatrix is obtained or until all combinations have been tested without success.

A first step 1000 sets the number of source packets Nb Source to 1. Then a step 1005 makes it possible to test whether this number is strictly lower than the total number of sources. If the number of source packets to be deleted is equal to the total source packet number, it means that no sub-matrix for source packet decoding exists. The system initializes the Rank variable to zero in a step 1065 and ends the algorithm in a step 1070. If the test of step 1005 is positive, the system proceeds to initialize two local variables Nb test and Iter in a step 1010. The variable Iter is initialized to the value 1, it is a counter which tests at each iteration a new set Nb Source and associated source packets to be deleted. The value Nb test corresponds to the number of sets of source packets to be deleted and is equal to the combination in the mathematical sense: Nb test = C (NbSource, Nb Source to be removed) = NbSource! / (Nb Source to be deleted)! (NbSource - No Source to delete)! It is to highlight that ! corresponds to the factorial mathematical operation. After the initializations, in a step 1015, the system chooses a set of sources to be suppressed among the sources, this set being different for each new iteration of Iter.

Then, in a step 1020, the system determines a sub-matrix excluding the received packets containing one of the sources of the set to be deleted. In a step 1025, the system then checks whether the matrix still has lines, that is to say that there remain received packets that do not contain sources of the set to be deleted. If there are no received packets, the system proceeds to step 1040. Otherwise, the system searches for the rank of the sub-matrix by the Gaussian pivot method in a step 1030. Then, the rank of the sub-matrix -matrix is compared to the number of source package remaining to be decoded in a step 1035. If the test is positive, it means that all the source packets present in the sub-matrix are decodable. The sub-matrix is then stored in a step 1055 and then the algorithm ends in the step 1060. If the test 1035 is negative, the system tests in a step 1040 if the value of the counter Iter is equal to the number of tests Nb test to perform, that is to say that all combinations have been made for a given number of source packages to delete. If the test of step 1040 is negative, the counter Iter is incremented by one unit and a new set of sources to be suppressed among the sources is determined in a step 1045. Then the step 1020 of construction of the sub-matrix is executed again. If the test 1040 is positive, then the system increases the number of sources to be deleted in a step 1050 and then the test of step 1005 is performed again. It will be noted that the invention is not limited to a purely material implantation but that it can also be implemented in the form of a sequence of instructions of a computer program or any form mixing a material part and a part software. In the case where the invention is partially or totally implemented in software form, the corresponding instruction sequence can be stored in a removable storage means (such as for example a floppy disk, a CD-ROM or a DVD-ROM) or no, this storage means being readable partially or totally by a computer or a microprocessor.

Claims (18)

REVENDICATIONS1. A network encoding method in a mesh network comprising at least one relay node which, in accordance with a predetermined network coding scheme, expects to receive a plurality of payload data packets and to output a predetermined resultant packet obtained in accordance with a a predetermined linear combination of said plurality of payload packets, each of the payload packets being formed, according to said predetermined network coding scheme, by a linear combination of one or more source packets, the source packets being transmitted by source nodes , said method being implemented by at least one relay node and being characterized in that it comprises the steps of: - detecting (610, 615, 630) if at least one input notification indicates that a packet useful is missing or has been received modified so that the linear combination of the source packet (s) forming said useful packet modified does not conform to the predetermined network coding scheme; in the case of a positive detection, constructing (635) a resulting packet modified from the received useful packets; constructing (640) an output notification identifying the differences between the linear combination of source packets representative of the predetermined resulting packet and the linear combination of source packets representative of the modified resulting packet; and - transmitting (650) the resulting modified packet and the output notification. 2. Method according to claim 1, characterized in that the input notification represents: an event internal to the relay node (510) when it indicates that a useful packet is missing; or - information received with a useful packet when it indicates that said useful packet has been received modified. 20 25 30 3. Method according to any one of claims 1 and 2, characterized in that each entry or exit notification comprises M fields, with M the number of source nodes, each of the M fields being associated with one of the M source packets generated by the M source nodes, each of the M fields comprising: a first subfield (415, 425) indicating whether a coefficient, associated with the source packet associated with said field, has been modified or not, in said linear combination of source packets representative of said modified resulting packet, with respect to said linear combination of source packets representative of said predetermined resulting packet; a second sub-field (420, 430) giving information relating to a value of the modified coefficient to which the first subfield relates, if the first subfield (415, 425) indicates a modification. 4. Method according to claim 3, characterized in that the second subfield (420, 430) is present only if the first subfield (415, 425) indicates a modification. 5. Method according to any one of claims 3 and 4, characterized in that the information included in the second subfield (420, 430) explicitly includes the value of the modified coefficient. 6. Method according to any one of claims 3 and 4, characterized in that the information included in the second subfield (420, 430) comprises at least one indication allowing a node processing said notification of entry or of exit to find the value of the modified coefficient. A method of network coding in a mesh network comprising at least one destination node which, according to a predetermined network coding scheme, expects to input a plurality of payload packets and is to output a plurality of source packets. decoded ones each obtained in a predetermined linear combination of said plurality of payload packets, each of the useful data packets being formed, according to said predetermined network coding scheme, by a linear combination of one or more source packets, the source packets being transmitted by source nodes, said method being implemented by at least one destination node and being characterized in that it comprises steps of: - detecting (910, 915, 930) if at least one input notification indicates that a useful packet is missing or has been received modified so that the linear combination of the source packet (s) forming said modified useful packet does not conform to the predetermined network coding scheme; in the case of a positive detection, obtaining (935, 940, 945, 950, 955, 975) a modified network coding scheme, as a function of the linear combinations of source packets forming the received useful packets; and decoding the received payload packets (960, 980) based on said modified network coding scheme to obtain at least one decoded source packet. 15 The method according to claim 7, characterized in that the step of obtaining a modified network coding scheme comprises the steps of: modifying (810, 815, 820, 920) a reference destination matrix to obtain a modified destination matrix, based on said at least one detected input notification; detecting (935, 940) whether the rank of said modified destination matrix is equal to the total number of source packets to be decoded; and - if the rank of said modified destination matrix is equal to the total number of source packets to be decoded, using (975, 980) said modified destination matrix to generate all decoded source packets in said step of decoding. The method of claim 8, characterized in that, if the rank of said modified destination matrix is not equal to the total number of source packets to be decoded, the step of obtaining a modified network encoding scheme includes 30 steps of: determining (945, 950) a sub-matrix of said modified destination matrix, whose rank is different from zero; using (955, 960) said sub-matrix to generate at least one decoded source packet, in said step of decoding. The method according to any one of claims 7 to 9, characterized in that the input notification represents: an event internal to the relay node when it indicates that a useful packet is missing; or - information received with a useful packet when it indicates that said useful packet has been received modified. 11. Method according to any one of claims 7 to 10, characterized in that each input notification comprises M fields, with M the number of source nodes, each of the M fields being associated with one of the M source packets generated. by the M source nodes, each of the M fields comprising: a first subfield (415, 425) indicating whether a coefficient of the source packet associated with said field has been modified or not, in said linear combination of source packets representative of a expected useful package; a second sub-field (420, 430) giving information relating to a value of the modified coefficient to which the first subfield relates, if the first subfield (415, 425) indicates a modification. The method of claim 11, characterized in that the second subfield (420, 430) is present only if the first subfield (415, 425) indicates a change. 13. Method according to any one of claims 11 and 12, characterized in that the information included in the second subfield (420, 430) explicitly includes the value of the modified coefficient. 14. Method according to any one of claims 11 and 12, characterized in that the information included in the second subfield (420, 430) comprises at least one indication allowing a node receiving said entry notification to find the value of the modified coefficient. 15. Computer program product downloadable from a communication network and / or recorded on a computer readable medium and / or executable by a processor, characterized in that it comprises program code instructions for the implementation of the Method according to at least one of Claims 1 to 6 and / or the method according to at least one of Claims 7 to 14, when said program is executed on a computer. Computer-readable storage medium, possibly totally or partially removable, storing a computer program comprising a set of instructions executable by a computer for carrying out the method according to at least one of claims 1 to 6 and / or the process according to at least one of claims 7 to 14. 17. A relay node included in a mesh network and participating in a network coding method in said mesh network, said relay node, according to a predetermined network coding scheme, expecting to receive as input a plurality of payload data packets and in front of outputting a predetermined resulting packet obtained according to a predetermined linear combination of said plurality of payload packets, each of the payload data packets being formed, according to said predetermined network coding scheme, by a linear combination of one or more source packets , the source packets being transmitted by source nodes, said relay node being characterized in that it comprises: detection means, making it possible to detect whether at least one input notification indicates that a useful packet is missing or that it was received modified so that the linear combination of the source packet (s) forming said useful packet is not in accordance with the predetermined network coding scheme; - first constructive means, activated in the case of positive detection, for constructing a modified result packet from the received useful packets; second construction means for constructing an output notification identifying the differences between the linear combination of source packets representative of the predetermined resulting packet and the linear combination of source packets representative of the modified resulting packet; transmission means for transmitting the modified resulting packet and the output notification. A destination node included in a mesh network and participating in a network coding method in said mesh network, said destination node, according to a predetermined network coding scheme, expecting to receive as input a plurality of payload data packets and outputting a plurality of decoded source packets each obtained in a predetermined linear combination of said plurality of payload packets, each of the data packets being formed, according to said predetermined network coding scheme, by a linear combination of one or more source packets, the source packets being transmitted by source nodes, said destination node being characterized in that it comprises: detection means, making it possible to detect whether at least one input notification indicates that a useful packet is missing or that it has been received modified so that the linear combination of the source packet or forms said modified useful packet does not conform to the predetermined network coding scheme; means of obtaining, activated in the case of a positive detection, making it possible to obtain a modified network coding scheme, as a function of the linear combinations of source packets forming the useful packets received; decoding means, for decoding the received useful data packets according to said modified network coding scheme, to obtain at least one decoded source packet. 25 30