JP5989346B2 - Data retransmission method and apparatus - Google Patents

Description

  The present invention relates to a network coding technique, and more particularly to a data retransmission method and apparatus.

  Network coding technology is an information exchange technology that combines routing and coding, and it is basically allowed that intermediate nodes in the network perform some degree of coding processing on received information flows. It is thought. Thereby, the network information flow can reach the limit of the maximum flow of multicast transmission, while the information flow cannot reach the limit of the maximum flow of multicast transmission in the traditional store and forward mode. In network coding, the information capacity of the network is fully utilized, the utilization rate of the existing network resources is greatly improved, and it is expected to be widely applied.

  Opportunistic network coding means that a node monitors and stores data packets broadcast from all neighboring nodes, and after the node receives the coded packets transmitted from an intermediate node, the cached data Using the packet, the decoding operation is performed on the encoded packet, and finally the original data packet is recovered.

  For broadcast characteristics of channels in wireless networks, opportunistic network coding has been applied to wireless transmission research, reducing the number of information transmissions, improving network throughput and energy utilization efficiency, reducing time delay, and transmission reliability. Used to improve Similarly, if the opportunistic network coding concept is applied to retransmissions in a wireless multicast network, the average number of retransmissions can be clearly reduced. One of the key problems to be solved in the wireless communication system is how to improve retransmission efficiency in view of the fact that the wireless channel has higher bit error rate and packet loss rate than the wired channel at the present stage. It is connected.

  In an embodiment of the present invention, a data retransmission method and apparatus are provided. This reduces the number of retransmissions and improves retransmission efficiency.

  In the data retransmission method submitted in the embodiment of the present invention, after the source node completes transmission of N data packets (N is a natural number), a packet loss pattern matrix of the N data packets is obtained. A and the source node analyzes the acquired packet loss pattern matrix to obtain a retransmission target data packet combination method B, and the source node converts the retransmission target data packet into the retransmission target data packet according to the combination method of the retransmission target data packet. And a step C of performing a combinational encoding to obtain an encoded packet to be retransmitted and sequentially retransmitting the encoded packet.

  Here, obtaining the packet loss pattern matrix of the N data packets means that the packet loss based on the ACK / NAK fed back from the terminal after the source node completes the transmission of the N data packets. Creating a pattern matrix.

  Specifically, every time the source node completes transmission of one data packet, all terminals use ACK / NAK to synchronously feed back their own packet loss status, and the source node After the transmission of the data packet is completed, a packet loss pattern matrix is created based on ACK / NAK fed back from all terminals. Alternatively, every time the source node completes transmission of one data packet, M terminals out of all terminals receiving the data packet use ACK / NAK to check the packet loss status of the local station. After synchronous feedback, the source node completes transmission of N data packets, and then creates a packet loss pattern matrix based on ACK / NAK fed back from the M terminals.

  In the above method, the source node selects M terminals having the worst signal quality among all the terminals based on the reference signal transmitted from the terminal as the terminal that feeds back the ACK / NAK. It may further include notifying the M terminals to feed back ACK / NAK via control signaling.

  The range of the preferable value of M is 10-15, and may be equal to 10, for example.

  The source node analyzes the acquired packet loss pattern matrix to obtain the data packet combination method to be retransmitted, by removing the zero-eliminated packet loss from the packet loss pattern matrix by removing the columns whose elements are all zero. Obtaining a pattern matrix and dividing the zero-removed packet loss pattern matrix into two sub-matrices, wherein a plurality of encoded packets corresponding to each column of the first sub-matrix can be retransmitted in combination On the other hand, the data packet corresponding to each column of the second sub-matrix is a data packet that cannot be retransmitted in combination with other data packets, and is divided into the first sub-matrix and the first sub-matrix after the division. The columns with the two sub-matrices are further merged to obtain a second packet loss pattern matrix and a third sub-matrix, where the merged second matrix Only one of the elements of the packet loss pattern matrix column is equal to 2, and the third sub-matrix column is merged with the second packet loss pattern matrix to reflect the method of combining the data packets to be retransmitted. A third packet loss pattern matrix, where all elements in the merged column are all 1 or less.

Removing a column whose elements are all zero from the packet loss pattern matrix extracts a column where all elements in the packet loss pattern matrix Ω are zero to form one sub-matrix ω ⁰ and the remaining Comprising a zero-removed packet loss pattern matrix ω ^α in columns.

Here, dividing the zero-removed packet loss pattern matrix into two sub-matrices selects columns of the zero-removed packet loss pattern matrix ω ^α in order, and the following selected from the zero-removed packet loss pattern matrix ω ^α : _Add all the columns that satisfy the condition (1) that has not been selected in other columns and the condition (2) that ∀i and R _i (θ _βk ) ≦ 1 after merging, A column θ _βk is generated, where R _i (θ _βk ) represents the value of the element in the i-th row of the column vector θ _βk (1 ≦ k ≦ K), and the data packet corresponding to the selected column is generated. On the other hand, combinatorial coding using modulo-2 addition of bits as a basic operation is performed to obtain a coded packet P _k, and the column after addition is put into the first sub-matrix ω ^β , and the column that cannot be merged is Into the sub-matrix ω ^γ .

、ことを含む。 Obtaining the second packet loss pattern matrix and the third sub-matrix by further merging the columns of the first sub-matrix and the second sub-matrix after the division of the second sub-matrix ω ^γ All columns are added to the end of the first sub-matrix ω ^β to form a second packet loss pattern matrix ω ^βγ ,

, Including that.

Merging the columns of the third sub-matrix with the second packet loss pattern matrix is e (i ^* ) = 1 for the column vector e of the third sub-matrix ω ^{μ and} the column of ω ^βγ θ _βγh (1 ≦ h ≦ H) is determined in order, and when R _{i *} (e∪θ _βγh ) = 1 is satisfied, it is determined that the target column of e is θ _βγh and e is added to the target column Here, R _{i *} (e∪θ _βγh ) = 1, which represents the sum of the values of the elements of the i ^* -th row of the column vector e and the column vector θ _βγS, and corresponds to the column vector e and incorporated into encoded packets _{P h (1 ≦ h ≦ H} ) corresponding to the target column theta _Betaganmah, the third sub-matrix omega ^mu, removing the column vector e, equal to the third sub-matrix omega ^mu Group the missing column vectors, add them, put the resulting column into the third packet loss pattern matrix, and Then, for each group, combination encoding is performed on the data packet corresponding to each column vector in the group to obtain an encoded packet.

  The source node performs the combination encoding on the retransmission target data packet according to the combination scheme of the retransmission target data packet to obtain the encoded packet of the retransmission target. The source node of the third packet loss pattern matrix In accordance with the combination scheme shown in the corresponding encoded packet or data packet for each column, the retransmission target data packet is subjected to combination encoding based on modulo-2 addition of bits as a basic operation, and the retransmission target encoded packet is Including.

The data retransmission method according to the embodiment of the present invention performs step B, determines whether there is a data packet that needs to be transmitted, and if there is a data packet that needs to be transmitted, N ′ = floor ( ΔT / T _S ) <N new data packets are continuously transmitted, where ΔT is necessary for the source node to analyze the packet loss pattern matrix and obtain a combination method of data packets to be retransmitted. T _S represents the time taken for the source node S to transmit one data packet, the floor (X) function represents the smallest integer greater than the rational number X, and executes step C After that, N−N ′ data packets are continuously transmitted to obtain a packet loss pattern matrix of the newly transmitted N data packets. Back-up to B, further comprising.

  The maximum number of retransmissions for the source node to retransmit the encoded packet is 5 or less.

  The data retransmission apparatus according to the embodiment of the present invention provides a packet loss pattern matrix acquisition unit that acquires a packet loss pattern matrix of N data packets every time transmission of N data packets (N is a natural number) is completed. And a combination method determining means for analyzing the acquired packet loss pattern matrix to obtain a combination method of retransmission target data packets, and performing combination encoding on the retransmission target data packet according to the combination method of retransmission target data packets. And an encoding means for obtaining an encoded packet to be retransmitted, and a retransmission means for sequentially retransmitting the encoded packet.

  Here, the combination method determining means removes a column in which all elements are zero from the packet loss pattern matrix to obtain a zero removal packet loss pattern matrix, and combines the zero removal packet loss pattern matrix into two sub-sequences. Divided into matrices, where the encoded packets corresponding to each column of the first sub-matrix are a plurality of data packets that can be retransmitted in combination, while each column of the second sub-matrix The data module corresponding to the data module is a data packet that cannot be retransmitted in combination with other data packets, and the columns of the divided first sub-matrix and second sub-matrix are further merged , To obtain a second packet loss pattern matrix and a third sub-matrix, where among the elements of the columns of the second packet loss pattern matrix after merging, An extension module in which at least one is equal to 2, and a third packet loss pattern matrix reflecting the combination scheme of retransmission target data packets by merging the columns of the third sub-matrix with the second packet loss pattern matrix Where all elements in the post-merger column are merge modules that are all 1 or less.

  The data retransmission method and apparatus according to an embodiment of the present invention can effectively combine retransmission target data packets by analyzing a packet loss pattern matrix to obtain a combination method of retransmission target data packets. This reduces the number of retransmissions and improves retransmission efficiency.

It is a figure which shows the topology model of a wireless multicast network. 3 is a flowchart of a data retransmission method according to an embodiment of the present invention. It is a figure which shows the structure of the data retransmission apparatus which concerns on the Example of this invention. When the number of terminals receiving data packets is 1000 and the data transmission method of this embodiment is adopted, the graphs each show the relationship between the decoding success probability and the number of retransmissions in the case of full feedback or partial feedback. is there. It is a graph which shows the relationship between the frequency | count of a part feedback, and a decoding success probability.

  In order to improve the retransmission efficiency of the wireless multicast network, a data retransmission method is provided in an embodiment of the present invention.

In this embodiment, the topology model of the wireless multicast network that implements the retransmission algorithm is as shown in FIG. In the wireless multicast network shown in FIG. 1, S represents a source node, T _i (1 ≦ i ≦ M) represents M terminals, and S represents N data packets P _j (1 ≦ j ≦). N) is broadcast, M and N are natural numbers, and the time slot that completes transmission (or retransmission) of one data packet is ST, during which all terminals T _i are ACK / Using NAK, the packet loss status of the own station is synchronously fed back, and the number i of the own station is notified to S. In the present invention, it is assumed that the transmission of ACK / NAK and retransmission packet is reliable, that is, there is no loss. After completing transmission of N data packets, S can obtain a packet loss pattern (Loss Pattern) of M terminals based on ACK / NAK information fed back from the M terminals. This packet loss pattern can be represented by one M × N matrix and is denoted as a packet loss pattern matrix Ω. The element β _ij of Ω represents whether or not the i th user has received the j th data packet, and the range of values of β _ij is {0, 1}, where β _ij is equal to 0 Represents successful reception, and β _ij equal to 1 represents packet loss. In Ω, the i-th row represents the packet loss status of the terminal T _i , and the j-th column represents the reception status of the data packet P _j . Table 1 below shows an example of the packet loss pattern matrix Ω. In Table 1, N = 10 and M = 5.

  For convenience of explanation, before describing specific embodiments of the present invention, first, the following definitions are given.

  Definition 1: A sub-matrix ω of a packet loss pattern matrix Ω having a size of M × n (n ≦ N). Here, M represents the number of terminals, N represents the number of data packets, and M, N ≧ 1.

Definition 2: R _i (ω) represents the sum of the values of all elements in the i-th row of the matrix ω, and therefore 0 ≦ R _i (ω) ≦ n (1 ≦ i ≦ M), and C _j (Ω) represents the sum of the values of all elements in the j-th column of ω, so that 0 ≦ C _j (ω) ≦ M (1 ≦ j ≦ n), and O (ω) This represents the number of R _i (ω) ≠ 0, whereby 0 ≦ O (ω) ≦ M can be obtained.

Definition 3: D _j (1 ≦ j ≦ N) represents the computational overhead for all terminals to recover the data packet P _j . Specifically, when a certain terminal recovers P _j , D _j is incremented by 1 every time an exclusive OR operation is performed.

When one terminal loses ε (1 ≦ ε ≦ N) data packets, the terminal obtains at most one data packet from each retransmission packet. Therefore, the retransmission algorithm requires at least ε retransmissions and at most N retransmissions. When M terminals have lost ε ₁ , ε ₂ ,..., _M data packets, respectively, when retransmission is performed according to the broadcast method, the number of retransmissions Rt of the retransmission algorithm is always the condition shown in Equation 1 below. Meet.

[Formula 1]
Max (ε _i ) <Rt ≦ Min (N, Σε _i ), i = 1, 2,..., M

  The main goal of the data retransmission method according to the embodiment of the present invention is to reduce the number of retransmissions Rt as much as possible. Hereinafter, embodiments of the present invention will be described in detail with reference to specific examples and drawings. As shown in FIG. 2, the data retransmission method provided by the embodiment of the present invention mainly includes the following steps.

  In step 1, the source node S acquires a packet loss pattern matrix Ω after completing transmission of N data packets.

  In the wireless multicast network, the terminal can feed back the packet loss status of the own station to the source node S by adopting a method of full feedback or partial feedback.

In the case of the all feedback method, every time the source node S completes transmission of one data packet, all terminals use ACK / NAK to synchronously feed back their own packet loss status and To the source node S. For example, when the data packet is correctly received, ACK is fed back, and when the data packet is not received or when there is a reception error, NAK is fed back. In this way, after completing the transmission of N data packets, the source node S can create a packet loss pattern matrix Ω based on ACK / NAK fed back from all terminals. For example, an M × N matrix is created, in which element β _ij indicates whether the i th user has received the j th data packet, and the range of values of β _ij is {0, 1} _Yes , where β _ij is equal to 0 indicates successful reception, and β _ij is equal to 1 indicates packet loss.

In the case of partial feedback, every time the source node S completes transmission of one data packet, some terminals in the wireless multicast network synchronize their own packet loss status using ACK / NAK. While feeding back and informing the source node S of the number i of the own station, the other terminals do not feed back the packet loss status of the own station to the source node S. Also in this case, after completing the transmission of N data packets, the source node S can create the packet loss pattern matrix Ω based on the ACK / NAK fed back from the some terminals. For example, similarly, an M × N matrix is created, and an element β _ij therein indicates whether the i-th user has received the j-th data packet, and the range of values of β _ij is {0, 1 }, Where β _ij equals 0 indicates successful reception, β _ij equals 1 indicates packet loss, and M feeds back its packet loss status to the source node S. Represents the number of terminals.

  In step 2, the source node S analyzes the packet loss pattern matrix Ω to obtain a combination method of retransmission target data. That is, the source node S analyzes the packet loss pattern matrix Ω to obtain a lost packet combination method when a retransmission packet is transmitted.

  In step 3, the source node S performs combination coding on the retransmission target data packet according to the combination scheme of the retransmission target data packet to obtain a retransmission target encoded packet, and sequentially retransmits the encoded packet.

  The combination encoding in this step is specifically combination encoding using a modulo-2 addition of bits as a basic operation.

  In the present embodiment, the source node S can effectively combine the retransmission target data packets by analyzing the packet loss pattern matrix Ω to obtain a combination scheme of the retransmission target data packets. This reduces the number of retransmissions and improves retransmission efficiency.

  In step 2 above, the source node S analyzes the packet loss pattern matrix Ω to obtain a combination method of data packets to be retransmitted. Specifically, the source node S has four processes: zero elimination, division, extension, and merger. Including. Hereinafter, each of these four processes will be described in detail.

  The first process of zero elimination is to remove the columns where all elements are zero from the packet loss pattern matrix Ω.

Since the occurrence probability of packet loss at M terminals is independent from each other, the position of an element having a value of 1 in the packet loss pattern matrix Ω is also completely random. Depending on the number of elements with a value of 1 in each column of Ω, Ω can be divided into two sub-matrices ω ⁰ and ω ^α . These two sub-matrices each have M rows and n ₀ and n _α columns, respectively _. ω ⁰ and ω ^α satisfy definition 1 and satisfy n ₀ + n _α = N. Here, ω ⁰ is a zero matrix and satisfies C _j (ω ⁰ ) = 0. ω ^α is composed of columns satisfying 0 <C _j (ω ^α ) ≦ M in Ω, and is called a zero-removed packet loss pattern matrix. That is, in the above process, a column in which all the elements are zero is extracted from the packet loss pattern matrix Ω to form one sub-matrix ω ⁰ , and the zero removal packet loss pattern matrix ω is formed from the remaining columns. ^α is constructed. If at least one of ω ⁰ and ω ^α is present and no matrix is present, the number of columns of the matrix is zero. Here, since the data packet corresponding to the sub-matrix ω ⁰ has already been correctly received by all the terminals, there is no need to retransmit. On the other hand, the data packet corresponding to the zero removal packet loss pattern matrix ω ^α needs to be retransmitted.

Taking the packet loss pattern matrix Ω shown in Table 1 as an example, ω ⁰ is the fifth column of Ω, and ω ^α is composed of the remaining nine columns other than the fifth column in Ω. Table 2 below shows a zero removal packet loss pattern matrix ω ^α obtained by performing zero removal processing on the packet loss pattern matrix Ω shown in Table 1.

The second process of division is to divide the zero removal packet loss pattern matrix ω ^α into two sub-matrices. Here, the encoded packet corresponding to each column of the first sub-matrix is a plurality of data packets that can be retransmitted in combination, while the data packet corresponding to each column of the second sub-matrix. Is a data packet that cannot be retransmitted in combination with other data packets. That is, by processing the columns of zero remover packet loss pattern matrix omega ^alpha, is divided into first and submatrix omega ^beta 2 one new packet loss pattern sub-matrix and second sub-matrix omega ^gamma. Here, the column θ _βk (1 ≦ k ≦ K) of the first sub-matrix ω ^β corresponds to the encoded packet P _k . The element η _ij of the first sub-matrix ω ^β represents the number of lost packets of the i-th user in the j-th encoded packet combination, and the set of values of η _ij is {0, 1}. The second sub-matrix ω ^γ is a sub-set of the zero removal packet loss pattern matrix ω ^α .

  Specifically, the above dividing process includes the following steps.

In the first step, the columns of the zero-removed packet loss pattern matrix ω ^α are sequentially selected, and the condition (1 that has not been selected in the following other columns selected from the zero-removed packet loss pattern matrix ω ^α) (1 ), ∀i, and condition (2) that R _i (θ _βk ) ≦ 1 after merging are added _together to generate a column θ _βk .

In the second step, the data packet corresponding to the selected column is subjected to combination encoding using a modulo-2 addition of bits as a basic operation to obtain an encoded packet P _k, and the column after the addition is Into one submatrix ω ^β .

In the third step, the columns that cannot be merged are placed in the second sub-matrix ω ^γ .

As should be explained, the encoded packet P _k obtained by the division can be completely decoded. The reason is that the encoded packet P _k includes at most one lost packet of the user, and the user can obtain the lost packet by performing a few exclusive OR operations.

For example, when dividing the zero remover packet loss pattern matrix omega ^alpha shown in Table 2 above, as an example the first row of omega ^alpha, adds the omega ^alpha select the P ₁ and P _2, the theta _.beta.1 Generate. Similarly, by adding the omega ^alpha select the P ₃ and P _4, generates theta _.beta.2, and adds the omega ^alpha Select P ₆ and P _8, is possible to generate theta _.beta.3 it can. Thereby, the first sub-matrix ω ^β is obtained. P _7, P ₉ and P ₁₀ of ω ^α are data packets that do not satisfy the addition condition and cannot be merged. Therefore, the columns corresponding to P _7, P ₉ and P ₁₀ are put in the second sub-matrix ω ^γ . Tables 3 and 4 below show a first sub-matrix ω ^β and a second sub-matrix ω ^γ obtained by dividing the zero-removed packet loss pattern matrix ω ^α shown in Table 2, respectively.

The third process of expansion, that is, further merging the columns of the first sub-matrix ω ^β and the second sub-matrix ω ^γ after the division into the second packet loss pattern matrix ω ^βγ and the third The submatrix ω ^μ is obtained. Here, at most one of the elements of the merged column is equal to 2. This expansion process specifically includes the following steps.

In the first step, all the columns of the second sub-matrix ω ^γ are added to the end of the first sub-matrix ω ^β to create a new packet loss pattern matrix called the second packet loss pattern matrix ω ^βγ. Configure.

In the second step ,

In the third step ,

The column of the second packet loss pattern matrix ω ^βγ obtained after the expansion is completed is θ _βγh (1 ≦ h ≦ H). For the terminal, the encoded packet at this stage may include two lost packets necessary for the local station. Therefore, the terminal cannot recover the lost packet.

In the process of expanding the first sub-matrix ω ^β and the second sub-matrix ω ^γ shown in Table 3 and Table 4 above, only P ₉ satisfies the expansion condition and is incorporated into the first encoded packet. Tables 5 and 6 below show the second packet loss pattern matrix ω ^βγ obtained by extending the first sub-matrix ω ^β and the second sub-matrix ω ^γ shown in Tables 3 and 4, and 3 sub-matrices ω ^μ are shown respectively.

The fourth process of merging is to merge the columns of the third sub-matrix ω ^μ with the second packet loss pattern matrix ω ^βγ to obtain a third packet loss pattern matrix. Here, all the elements of the merged column are all 1 or less. The purpose of the merger is to solve the problem that terminals due to expansion cannot recover lost packets.

  The merger process specifically includes the following steps.

In the first step, e (i ^* ) = 1 for the column vector e of ω ^μ , the column θ _{βγh of} ω ^βγ is judged in order, and R _{i *} (e∪θ _βγh ) = 1 is satisfied. In this case, it is determined that the target column of e is θ _βγh , and e is added to the target column.

In the second step, the data packet corresponding to e is incorporated into the encoded packet P _h (1 ≦ h ≦ H) corresponding to the target column θ _βγh , and the column vector e is obtained from the third sub-matrix ω ^μ. Remove.

In a third step, the unequal column vectors in the third sub-matrix ω ^μ are grouped and added, the resulting columns are placed in a third packet loss pattern matrix, and for each group, the group The data packet corresponding to each column vector is subjected to combination coding to obtain a coded packet.

When the merge processing is performed on the second packet loss pattern matrix ω ^βγ and the third sub-matrix ω ^μ shown in Tables 5 and 6, the target column of the column vector e corresponding to the expanded P ₉ is θ since at _.beta.2, with merging e to theta _.beta.2, incorporated into the encoding packets corresponding to P ₉ to theta _.beta.2. Table 7 below shows a third packet loss pattern matrix obtained by merging the second packet loss pattern matrix ω ^βγ and the third sub-matrix ω ^μ shown in Tables 5 and 6.

  The third packet loss pattern matrix reflects the combination method of retransmission target data packets, that is, the encoded packet or data packet corresponding to each column of the third packet loss pattern matrix is a retransmission target data packet. Reflects the combination method.

  When the third packet loss pattern matrix is obtained, in step 3, the source node S performs retransmission according to the combination scheme shown in the encoded packet or data packet corresponding to each column of the third packet loss pattern matrix. It is possible to obtain a coded packet to be retransmitted by performing combination coding using modulo-2 addition of bits as a basic operation on the data packet, and sequentially retransmit the coded packet.

並びにデータパケットＰ_７およびＰ_１０に、改めて組み合わせることができる。このようにして、５回の再送だけで、９個の再送対象データパケットの再送を完成することができる。具体的に、端末Ｔ_３は、１番目の再送パケットを受信した時、

を回復することができ、次に、２番目の再送パケットから、Ｐ_９を回復して、もう一度排他的論理和処理を行うと、Ｐ_１を得ることができる。一方、ほかのユーザにとって、すべての符号化パケットは、完全に復号化できるものである。ここから分かるように、ゼロ除去、分割、拡張および合併という４つの過程により得られた再送対象データパケットの組み合わせ方式では、データパケットの再送回数を大幅に減少できるとともに、端末によるデータパケットに対する正しい回復を保証できる。 As can be seen from the above method, the source node S can obtain a data packet combination method to be retransmitted through four processes of zero elimination, division, extension, and merger, and can greatly reduce the number of retransmissions of data packets. At the same time, correct recovery of the data packet by the terminal can be guaranteed. In particular, when the number of user terminals is relatively large but the number of data packets is relatively small, the effect is further clarified. For example, corresponding to the packet loss pattern matrix Ω shown in Table 1, data packets P ₁ , P ₂ , P ₃ , P ₄ , P ₆ , P ₇ , P that need to be retransmitted by the processing of the above four processes. ₈ , P ₉ and P ₁₀ ,

In addition, the data packets P ₇ and P ₁₀ can be combined again. In this way, retransmission of nine retransmission target data packets can be completed with only five retransmissions. Specifically, when the terminal T ₃ receives the first retransmission packet,

Next, P ₉ is recovered from the second retransmission packet, and exclusive OR processing is performed once again to obtain P ₁ . On the other hand, for other users, all encoded packets can be completely decoded. As can be seen from the above, in the combination method of data packets to be retransmitted obtained through the four processes of zero elimination, division, extension and merger, the number of retransmissions of data packets can be greatly reduced, and correct recovery of data packets by the terminal can be achieved. Can guarantee.

The following content is pseudo code in which the source node S divides, extends, and merges the packet loss pattern matrix Ω.

In addition, as described above, in the above data retransmission method, after obtaining the packet loss pattern matrix Ω of N data packets, the source node S obtains a method of combining retransmission packets by analyzing Ω. It is necessary to start retransmission. Therefore, there is a certain time delay from the acquisition of the packet loss pattern matrix Ω to the start of retransmission. Assuming that this time delay is ΔT, ΔT is a processing time required for the source node S to analyze the packet loss pattern matrix Ω and obtain a combination method of data packets to be retransmitted. In order not to affect the data transmission efficiency of the system, perform step 2 above, determine if there are new data packets that need to be transmitted, and if there are new data packets that need to be transmitted, N ′ = Floor (ΔT / T _S ) <N new data packets are continuously transmitted, where T _S represents the time taken for the source node S to send one data packet and floor (X) The function represents taking the smallest integer greater than the rational number X, and then performs step 3 above, completes the execution of step 3, and then continues to transmit N−N ′ data packets to create a new It is also possible to return to step 2 after obtaining the packet loss pattern matrix Ω of N data packets.

  As described above, the source node processes Ω and continues to transmit the next group of data packets. The source node can transmit data continuously for the remaining time, except that it causes interruption of data transmission during the last analysis on the packet loss pattern matrix Ω. This ensures the data transmission efficiency of the system.

  Corresponding to the above data retransmission method, a data retransmission apparatus is also provided in the embodiment of the present invention. As shown in FIG. 3, each time the apparatus completes transmission of N data packets, the apparatus acquires packet loss pattern matrix acquisition means for acquiring a packet loss pattern matrix of the N data packets, and acquired packets. The combination method determining means for analyzing the loss pattern matrix to obtain the combination method of the retransmission target data packet and the combination encoding for the retransmission target data packet according to the combination method of the retransmission target data packet, Encoding means for obtaining an encoded packet, and retransmission means for retransmitting the encoded packet in order.

  Here, the combination method determination means includes a zero removal module that removes columns whose elements are all zero from the packet loss pattern matrix to obtain a zero removal packet loss pattern matrix, and two zero removal packet loss pattern matrices. Divided into two sub-matrices, where the corresponding encoded packet for each column of the first sub-matrix is a plurality of data packets that can be retransmitted in combination, while each of the second sub-matrix The data module corresponding to each column is a data packet that cannot be retransmitted in combination with other data packets, and the columns of the divided first sub-matrix and second sub-matrix are further merged To obtain a second packet loss pattern matrix and a third sub-matrix, where at most one of the elements of the merged column is equal to 2. The extension module and the third sub-matrix column are merged with the second packet loss pattern matrix to obtain a third packet loss pattern matrix that reflects the combination scheme of the data packets to be retransmitted. All elements in the row include merged modules that are all 1 or less.

  Specifically, the zero elimination module, the division module, the expansion module, and the merge module described above complete the analysis and processing for the packet loss pattern matrix Ω by performing the processes of zero elimination, division, expansion, and merge respectively. To do.

  Similarly, in this embodiment, the data retransmission apparatus can effectively combine the data packets to be retransmitted by the processes of the zero removal module, the division module, the expansion module, and the merge module, greatly increasing the number of retransmissions of the data packet. It can be reduced and correct recovery of data packets by the terminal can be guaranteed.

  As should be explained, the data retransmission method and apparatus provided in the embodiments of the present invention are based on the packet loss pattern matrix Ω. Therefore, in the wireless multicast network, it is required that the packet loss information of the own station is fed back from the terminal to the source node S by adopting a full feedback or partial feedback method. Therefore, it is not applied to data retransmission in a wireless multicast network without feedback.

  As described above, in step 1 above, the source node S obtains the packet loss pattern matrix Ω based on ACK / NAK fed back from some terminals after completing transmission of N data packets. can do. Then, in the case of partial feedback, how to determine the number of terminals that feed back ACK / NAK is one of the problems to be solved in the case of partial feedback. The following conclusions can be obtained from a large amount of simulation. That is, regardless of the number of terminals that actually receive data packets, if the number of terminals that feed back ACK / NAK is set to 10 or about 10, a better decoding success probability can be obtained. In this case, simply increasing the number of terminals that feed back ACK / NAK does not significantly increase the probability of successful decoding, but conversely increases the computational complexity and overhead of the decoding and retransmission processes. Become.

  Also, how to select a terminal that feeds back ACK / NAK is one of the problems to be solved in the case of partial feedback. In this embodiment, it is possible to adopt a rule of feedback on the worst channel. That is, the node having the worst channel condition is fed back the reception status of the data packet of its own station. In actual application, first, a terminal receiving a data packet can be requested to transmit a reference signal to the source node. Next, based on the reference signal transmitted from each terminal, the source node selects, from among the terminals, M terminals having the worst signal quality as terminals that feed back ACK / NAK, and performs downlink control signaling. Via, the M terminals are notified that ACK / NAK needs to be fed back. In this way, when receiving the ACK / NAK fed back from the M terminals, the source node can generate the packet loss pattern matrix Ω.

  Next, based on the result of a large amount of simulation, it can be determined that it is relatively appropriate to set the maximum number of retransmissions to 5 or less when the source node retransmits the data packet. As described above, the request (for example, greater than 95%) regarding the probability of successful decoding by the user can be satisfied, and the overhead required for retransmission is within a controllable range. That is, if the number of retransmissions is simply increased after the number of retransmissions reaches 5, the retransmission overhead is significantly increased without significantly increasing the decoding success probability.

  FIG. 4 shows a case where the number of terminals receiving data packets is 1000, and when the data transmission method of this embodiment is adopted, all feedback or partial feedback (10 users feed back ACK / NAK). The relationship between the decoding success probability and the number of retransmissions in FIG. Here, the horizontal axis in FIG. 4 represents the number of retransmissions, the vertical axis represents the decoding success probability, and the curve indicated by the solid line in FIG. 4 represents the decoding success probability when all feedback is adopted. Representing the relationship with the number of retransmissions, the crossed curve in FIG. 4 represents the relationship between the decoding success probability and the number of retransmissions when a partial feedback is employed. As can be seen from FIG. 4, when the number of retransmissions is 5 or less, the decoding success probability obtained when partial feedback is adopted is higher than the decoding success probability obtained when all feedback is adopted. . That is, the algorithm according to the present embodiment is further applied to a case where a part of feedback is adopted because the maximum number of retransmissions is determined.

  As can be seen from the simulation results of FIG. 4, when partial feedback is employed, if the channel condition is relatively bad (packet loss rate is 0.1), the decoding success probability reaches 99%. I can't. For users with relatively high demands on video, such video quality cannot be tolerated. In order to further improve the video quality, the present invention proposes the concept of partial feedback multiple times. That is, based on the first feedback result, after completing one retransmission cycle, the ten worst users can be reselected to perform feedback, and then another retransmission cycle is performed. As shown in FIG. 5, when such a simple extension is adopted, the decoding success probability can reach 99% or more. FIG. 5 shows the relationship between the number of partial feedbacks and the decoding success probability. As can be seen from FIG. 5, the probability of successful decoding increases as the number of partial feedbacks increases.

  The above are only preferred embodiments of the present invention and do not limit the present invention. Various modifications, equivalent replacements, improvements and the like made within the spirit and principle of the present invention should all be included in the protection scope of the present invention.

Claims (14)

A data retransmission method,
The source node completes transmission of N (N is a natural number) data packets, and then obtains a packet loss pattern matrix of the N data packets.
The source node analyzes the acquired packet loss pattern matrix to obtain a combination method of data packets to be retransmitted, step B;
The source node performs combination coding on the retransmission target data packet according to a combination scheme of the retransmission target data packet to obtain a retransmission target encoded packet, and retransmits the encoded packet in order, step C;
Only including,
The source node analyzes the acquired packet loss pattern matrix to obtain a combination method of data packets to be retransmitted.
From the packet loss pattern matrix, remove all columns whose elements are all zero to obtain a zero removal packet loss pattern matrix,
Dividing the zero-removed packet loss pattern matrix into two sub-matrices, where the encoded packets corresponding to each column of the first sub-matrix are a plurality of data packets that can be retransmitted in combination The data packet corresponding to each column of the second sub-matrix is a data packet that cannot be retransmitted in combination with other data packets;
The columns of the first sub-matrix and the second sub-matrix after the division are further merged to obtain the second packet loss pattern matrix and the third sub-matrix, where the second packet after the merging At most one of the elements of the column of the loss pattern matrix is equal to 2,
The third sub-matrix column is merged with the second packet loss pattern matrix to obtain a third packet loss pattern matrix reflecting the combination scheme of the data packets to be retransmitted. Here, all the columns of the merged columns are obtained. The elements are all 1 or less,
Wherein the free Mukoto that. Obtaining the packet loss pattern matrix of N data packets is
After completing the transmission of N data packets, the source node creates a packet loss pattern matrix based on the ACK / NAK fed back from the terminal.
The method of claim 1, comprising: After completing the transmission of N data packets, the source node creates a packet loss pattern matrix based on ACK / NAK fed back from the terminal.
Every time the source node completes transmission of one data packet, all terminals use ACK / NAK to synchronously feed back their own packet loss status,
After the source node completes transmission of N data packets, it creates a packet loss pattern matrix based on ACK / NAK fed back from all terminals.
The method of claim 2, comprising: After completing the transmission of N data packets, the source node creates a packet loss pattern matrix based on ACK / NAK fed back from the terminal.
Each time the source node completes transmission of one data packet, M terminals out of all terminals receiving the data packet use ACK / NAK to synchronously feedback the packet loss status of the local station. And
After the source node completes transmission of N data packets, it creates a packet loss pattern matrix based on ACK / NAK fed back from the M terminals.
The method of claim 2, comprising: Based on the reference signal transmitted from the terminal, the source node selects M terminals having the worst signal quality among all the terminals as terminals to which ACK / NAK is fed back, via downlink control signaling Informing the M terminals to feed back ACK / NAK.
5. The method of claim 4, further comprising: The range of the value of M is 10-15.
6. A method according to claim 4 or 5, characterized in that Removing a column whose elements are all zero from the packet loss pattern matrix
Extract a column in which all elements in the packet loss pattern matrix Ω are zero to form one sub-matrix ω ⁰ and configure a zero-removed packet loss pattern matrix ω ^α with the remaining columns.
The method of claim 1 , comprising: Dividing the zero-removed packet loss pattern matrix into two sub-matrices is
Select the columns of the zero-removed packet loss pattern matrix ω ^α in order, and the following selected from the zero-removed packet loss pattern matrix ω ^α ,
Condition (1) that it has never been selected in another column,
∀i, condition (2) that R _i (θ _βk ) ≦ 1 after merging,
All columns that satisfy are added to generate a column θ _βk , where R _i (θ _βk ) represents the value of the i-th element of the column vector θ _βk (1 ≦ k ≦ K),
The data packet corresponding to the selected column is subjected to combination coding using modulo-2 addition of bits as a basic operation to obtain a coded packet P _k, and the column after the addition is _defined as the first sub-matrix ω ^β put in,
Put columns that cannot be merged into the second sub-matrix ω ^γ ,
The method of claim 1 , comprising: Further merging the columns of the first sub-matrix and the second sub-matrix after the division to obtain the second packet loss pattern matrix and the third sub-matrix,
All columns of the second sub-matrix ω ^γ are added to the end of the first sub-matrix ω ^β to form a second packet loss pattern matrix ω ^βγ ,

The method according to claim 1, characterized in that it comprises a call. Merging the columns of the third sub-matrix into the second packet loss pattern matrix
For the column vector e of the third sub-matrix ω ^μ , e (i ^* ) = 1, and the column θ _βγh (1 ≦ h ≦ H) of ω ^βγ is determined in order, and R _{i *} (e∪θ _{If βγh} ) = 1, it is determined that the target column of e is θ _βγh and e is added to the target column, where R _{i *} (e∪θ _βγh ) = 1 and the column vector e represents the sum of the values of the elements of the i ^{* th} row of e and the column vector θ _βγS ,
The data packet corresponding to the column vector e is incorporated into the encoded packet P _h (1 ≦ h ≦ H) corresponding to the target column θ _βγh , and the column vector e is removed from the third sub-matrix ω ^μ .
Group and add unequal column vectors in the third sub-matrix ω ^μ , put the resulting columns into a third packet loss pattern matrix, and for each group to each column vector in the group The corresponding data packet is subjected to combination encoding to obtain an encoded packet.
The method of claim 1 , comprising: The source node performs the combination encoding on the retransmission target data packet according to the combination method of the retransmission target data packet to obtain the encoded packet of the retransmission target,
A combination code in which the source node uses a modulo-2 addition of bits as a basic operation for a data packet to be retransmitted in accordance with a combination method shown in an encoded packet or a data packet corresponding to each column of the third packet loss pattern matrix To obtain the encoded packet to be retransmitted,
The method of claim 1 , comprising: Perform step B and determine if there are data packets that need to be transmitted and if there are data packets that need to be transmitted, N ′ = floor (ΔT / T _S ) <N new data The packet is transmitted continuously, where ΔT is a processing time required for the source node to analyze the packet loss pattern matrix and obtain a combination method of data packets to be retransmitted, and T _S is the source node Represents the time taken for S to send one data packet, and the floor (X) function represents the smallest integer greater than the rational number X,
After performing Step C, NN ′ data packets are continuously transmitted to obtain a packet loss pattern matrix of the newly transmitted N data packets, and then the process returns to Step B.
The method of claim 1 further comprising: The maximum number of retransmissions for the source node to retransmit the encoded packet is 5 or less,
The method of claim 1 further comprising: A data retransmission device,
Packet loss pattern matrix obtaining means for obtaining a packet loss pattern matrix of N data packets every time transmission of N data packets (N is a natural number) is completed;
Analyzing the acquired packet loss pattern matrix to obtain a combination method of retransmission target data packets;
According to a combination method of retransmission target data packets, encoding means for performing retransmission encoding on the retransmission target data packets to obtain encoded packets of retransmission targets;
Resending means for retransmitting the encoded packets in order;
Only including,
The combination method determining means includes
A zero elimination module that removes all zero elements from the packet loss pattern matrix to obtain a zero elimination packet loss pattern matrix;
Dividing the zero-removed packet loss pattern matrix into two sub-matrices, where the encoded packets corresponding to each column of the first sub-matrix are a plurality of data packets that can be retransmitted in combination A segmentation module in which a data packet corresponding to each column of the second sub-matrix is a data packet that cannot be retransmitted in combination with other data packets;
The columns of the first sub-matrix and the second sub-matrix after the division are further merged to obtain the second packet loss pattern matrix and the third sub-matrix, where the second packet after the merging An extension module in which at most one of the elements of the column of the loss pattern matrix is equal to 2,
The third sub-matrix column is merged with the second packet loss pattern matrix to obtain a third packet loss pattern matrix reflecting the combination scheme of the data packets to be retransmitted. Here, all the columns of the merged columns are obtained. A merged module whose elements are all 1 or less,
And wherein the free Mukoto a.

Images