A retransmission method based on immediately decomposable network coding under imperfect feedback

Abstract

The invention belongs to the technical field of wireless communication, and relates to a retransmission method based on immediately decomposable network coding under imperfect feedback; the method comprises constructing an imperfect feedback environment system, a source node broadcasts N source data packets to all sink nodes, and each sink node constructs a feedback data packet including a request data packet set; the source node constructs a state feedback matrix according to the feedback data packet, creates a trust point set and an untrust point set, and respectively adopts a trust state update strategy and a likelihood estimation strategy to calculate the state of the source data packet; the state feedback matrix is ​​updated, an immediately decomposable network coding graph is established, and the weight value of each vertex thereof is calculated; a coding packet is obtained by using a maximum weight clique search algorithm and broadcasted to a sink node; the sink node updates the feedback data packet and sends it to the source node; the invention can effectively reduce the completion delay and decoding delay of a receiving end, optimize the retransmission recovery process of a wireless multicast network, and reduce the number of retransmissions.

Description

Technical Field

[0001] The invention belongs to the technical field of wireless communications and relates to a retransmission method based on immediately decomposable network coding under imperfect feedback. Background Art

[0002] With the development of wireless communication technology, wireless communication services have comprehensively and deeply influenced and changed people's production and life, bringing more convenience to people. At the same time, wireless multicast networks usually involve source nodes sending multiple data packets to multiple destination nodes, and the destination nodes receive all or part of the data packets. However, due to the inherent losses of wireless media, such as multipath fading and interference, the wireless channel has an erasure characteristic, which leads to data loss at the destination node. If the lost packets cannot be effectively recovered through retransmission, it may cause a large delay constraint and reduce network performance.

[0003] When traditional reliable transmission technology recovers lost data packets of destination nodes, one retransmission can only recover one same lost packet of multiple destination nodes. The efficiency of retransmission recovery is low, thus wasting network resources. Network Coding (NC) was first proposed by R. Ahlswede et al. in 2000. It is mainly used to improve the throughput of multicast networks and solve the problem that actual network transmission cannot reach the maximum theoretical capacity of "maximum flow-minimum cut". The core idea of ​​network coding is to break through the limitation that only intermediate nodes are allowed to forward data in traditional networks, but to allow intermediate nodes to combine and encode the received data packets according to certain rules, and the destination node decodes the required data packets according to the corresponding rules. The idea of ​​network coding changes the "store-and-forward" mode of intermediate nodes. Network coding requires intermediate nodes to have strong data processing capabilities and can encode and combine multiple data into coded packets before forwarding. It effectively solves the conflicts caused by multiple data streams on the same link and improves the bandwidth utilization of the network.

[0004] Instantly decodable network coding can encode and decode data packets by using simple binary XOR operations, thereby achieving fast encoding and decoding, which can meet the low latency requirements of delay-sensitive networks. Instantly decodable network coding can be divided into two categories: strict instantly decodable network coding (S-IDNC) and generalized instantly decodable network coding (G-IDNC). Most IDNC coding technologies are considered under perfect feedback environments, but in actual network transmission, not only the forward link from the sender to the receiver will lose data information, but also the backward link from the receiver to the sender will lose feedback information due to inherent fading of the channel and various other interference factors. The commonly used strategy is to use maximum likelihood estimation to determine the actual receiving state of the receiver, but this strategy usually only judges whether the data packet is correctly received based on the packet loss rate, so the quality of the link will affect the judgment result, with a high misjudgment rate. Another strategy is to establish a data packet receiving state model with the help of a partially observable Markov decision process. Compared with the maximum likelihood estimation, it can more accurately determine the actual receiving state, but this strategy requires multiple iterative calculations and consumes more resources. Generally speaking, there are two main measurement indicators for the research of immediately decomposable network coding. One is the completion time, which refers to the total number of coding transmissions when all receiving ends receive all their lost data packets. The second is the decoding delay. For a single receiving end, when the received data coding packet cannot be immediately decomposable, the decoding delay will increase. The vertex weights in the IDNC graph are determined based on the minimization of the completion delay or the decoding delay, and then the maximum weight search greedy algorithm is used to select the retransmitted coding packet combination. The completion time and decoding delay performance indicators are usually considered separately, but the decoding delay is closely related to the completion time. Therefore, there is an urgent need for an immediately decomposable network coding retransmission method that comprehensively considers the completion time and decoding delay performance indicators under imperfect feedback network transmission to improve retransmission recovery and give full play to the advantages of immediately decomposable network coding. Summary of the invention

[0005] In order to solve the problem of how to comprehensively improve the performance of completion delay and decoding delay and optimize retransmission recovery under imperfect feedback of current immediately decomposable network coding, the present invention proposes a retransmission method based on immediately decomposable network coding under imperfect feedback, which is characterized by comprising the following steps:

[0006] S1. Using POMDP theory to construct an imperfect feedback environment system, the imperfect feedback environment system includes a network transmission model and a POMDP model, the network transmission model includes a source node and a sink node;

[0007] S2. The source node broadcasts N source data packets to all sink nodes, and each sink node constructs a feedback data packet for the reception of the source data packet; the feedback data packet includes a set of received data packets, a set of lost data packets, a set of request data packets, and a set of unknown state data packets;

[0008] S3. Each sink node sends a feedback packet to the source node to feedback its own source packet reception situation, and the source node constructs a state feedback matrix based on the source packet reception situation of each sink node feedback;

[0009] S4. Create a set of trusted points and a set of untrusted points according to the request data packet set of each sink node, and use a trust state update strategy and a likelihood estimation strategy to calculate the source data packet states in the trusted point set and the untrusted point set respectively;

[0010] S5. Update the state feedback matrix, establish an immediately solvable network coding graph and calculate the weight value of each vertex;

[0011] S6. The source node uses the maximum weight clique search algorithm based on the weight values ​​of the vertices to obtain a coded packet and broadcasts it to all sink nodes; the sink node updates the feedback data packet based on the coded packet and sends it to the source node;

[0012] S7. Determine whether all destination nodes have completed the retransmission of the source data packet. If so, end the process; if not, return to step S4.

[0013] Furthermore, the network transmission model includes a source node and multiple sink nodes. The source node broadcasts N source data packets to all sink nodes through a sending channel. Each sink node receives all or part of the source data packets according to its own needs, and sends feedback data packets through a feedback channel. The sending channel and the feedback channel are both deletion channels.

[0014] Furthermore, the POMDP theory is used to construct a POMDP model, which is defined as a six-tuple including system state, action, observation state, system state transition probability, system observation state transition probability and system immediate reward function, where:

[0015] The system state set S = {s1, s2, s3, s4...} represents a set of implicit finite states of a system, and the element s i Indicates the i-th possible system state of all sink nodes under the current coded packet reception condition;

[0016] Action set A = element a in {a1,a2,a3,a4...} i Indicates the i-th type of coded packet that the source node can send in the current state;

[0017] Observation state set Z = {z i ,z2,z3,z4...} i It indicates the possible reception situation of the source data packet observed by the source node according to the feedback information of all the sink nodes;

[0018] The system state transition probability T(s,a,s') represents the probability that the system will transition to state s' after the source node broadcasts the coded packet a to all sink nodes when the system is in state s.

[0019] The system observed state transition probability O(s',a,z) represents the probability that the source node broadcasts the coded packet a to all the sink nodes, and after the system transitions to state s', the sink node observes the probability that the coded packet receiving state is z;

[0020] The system immediate reward function V(s,a,s',z) represents the reward that can be obtained when action a is taken in the system state s and the current observed state is z, and the system state becomes.

[0021] Furthermore, the feedback data packet of each sink node includes four data packet sets:

[0022] Received data packet set H i :Sink node R i The set of successfully received source data packets, including primary data packets and secondary data packets;

[0023] The set of lost packets L i :Sink node R i The set of unwanted source packets that were not received;

[0024] Request packet set W i :Sink node R i The set of source packets that were intended to be received but were lost;

[0025] The set of unknown packets U i , Sink Node R i The set of source packets for which feedback information is lost.

[0026] Furthermore, the process of step S4 includes:

[0027] S41. Calculate the average request set according to the size of the request data packet set in all sink node feedback data packets, place the sink nodes corresponding to the request data packet set larger than the average request set into the trusted point set, and place the sink nodes corresponding to the request data packet set not larger than the average request set into the untrusted point set;

[0028] S42. The source data packet reception conditions of all sink nodes in the confidence point set are arranged and combined, and the confidence degree corresponding to all combinations is calculated using the confidence formula. The confidence formula is expressed as:

[0029]

[0030] Wherein, b(s') represents the confidence level when the current time slot state is s', τ(b,a,z) represents the probability of reaching state z after sending coded packet a in the state corresponding to the maximum confidence b(s) of the previous time slot; b(s) represents the confidence level when the previous time slot state is s, O(s',a,z) represents the probability that the source node broadcasts coded packet a to all sink nodes, and after the system is transferred to state s', the sink node observes that the coded packet receiving state is z; T(s,a,s') represents the probability that the system is transferred to state s' after the source node broadcasts coded packet a to all sink nodes in state s; S represents the system state set;

[0031] S43. Select the combination corresponding to the maximum confidence, and use the source data packet reception status as the current source data packet reception status of each sink node in the confidence point set, and update the state feedback matrix;

[0032] S44. Use a likelihood estimation strategy to calculate the source data packet reception status of the destination node in the non-confident point set, and update the state feedback matrix.

[0033] Furthermore, in step S5, a main graph and a secondary graph are constructed based on the updated state feedback matrix, wherein the main graph is formed by source data packets in the set of request data packets of all destination nodes; and the secondary graph is formed by source data packets in the set of lost data packets of all destination nodes; both the main graph and the secondary graph are vertex sets; the vertices in the main graph and the secondary graph are connected according to the edge generation conditions to generate an immediately decomposable network coding graph.

[0034] Furthermore, assuming that there is ij and ν kl Two vertices, vertex ν ij Represents the sink node R i The jth source packet among N source packets is lost, and vertex ν ij and ν kl If any of the following edge generation conditions are met, an edge is generated between two vertices to connect them. The edge generation conditions include:

[0035] 1.j=l, i.e., the destination node R i and the sink node R k Request retransmission of the same source packet;

[0036] 2.j∈H k And l∈Hi , that is, the sink node R i and the sink node R k The source data packets requested for retransmission belong to each other's received data packet set.

[0037] Furthermore, the weight value calculation formula of each vertex in the immediately solvable network coding graph is:

[0038]

[0039]

[0040]

[0041] Among them, a i,k represents vertex ν ij and vertex ν kl The connection relationship, ω ij represents vertex ν ij The weight of i Represents the sink node R i The minimum completion delay, W i Represents the sink node R i The request packet set, p i Represents the sink node R i The probability of source data packet loss in the forward link is called the forward link; γ represents a variable with an increase multiple.

[0042] Beneficial effects of the present invention:

[0043] The present invention proposes a retransmission method based on immediately decomposable network coding under imperfect feedback, which combines the confidence state update strategy and the likelihood estimation strategy to judge the data packet reception state, and proposes a new weight value by minimizing the completion delay, designs an optimized weight group search algorithm, and optimizes the selection and generation of the coded packet. Through a certain calculation cost, the network reception situation is inferred, which can effectively reduce the completion delay and decoding delay of the receiving end, optimize the retransmission recovery process of the wireless multicast network, and reduce the number of retransmissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A wireless multicast network transmission model of the present invention;

[0045] Figure 2 is a flow chart of the method of the present invention;

[0046] Figure 3 This is an immediately decomposable network coding graph according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] The present invention provides a retransmission method based on immediately decomposable network coding under imperfect feedback, such as Figure 2 As shown, the following steps are included:

[0049] S1. The POMDP theory is used to construct an imperfect feedback environment system, wherein the imperfect feedback environment system includes a network transmission model and a POMDP model, and the network transmission model includes a source node and a sink node.

[0050] Specifically, the network model of the present invention is as follows: Figure 1 As shown in the figure, there are two types of nodes in this model. One type is the source node, which is the node that sends the data packet and can also be called the sender. Figure 1 B; one type is the destination node, that is, the node that receives the data packet, which can also be called the receiving end, such as Figure 1 In this embodiment, M sink nodes are used, represented by {R1, R2, ..., R i ,...,R M}.

[0051] S2. The source node broadcasts N source data packets to all destination nodes, each destination node receives N source data packets or part of N source data packets, and constructs corresponding feedback data packets; the feedback data packets include a received data packet set, a lost data packet set, a requested data packet set, and a data packet set with an unknown status.

[0052] Specifically, the transmission process of the network model of the present invention is divided into two stages: the initial transmission stage and the retransmission recovery stage. Figure 1 As shown, in the initial transmission phase, the source node B sends N source data packets {P1, P2, ..., P N}Broadcast to all sink nodes, each sink node receives all source data packets or part of the source data packets according to its own needs, and sends feedback data packets through the feedback channel; the sending channel and the feedback channel are both deletion channels.

[0053] Furthermore, for any destination node, the source data packet that it wants to receive is called a primary data packet, and the source data packet that it does not want to receive is called a secondary data packet.

[0054] Specifically, after the initial transmission phase is completed, each destination node constructs a corresponding feedback data packet; the feedback data packet includes four data packet sets:

[0055] Received data packet set H i :Sink node R i The set of successfully received source data packets, including primary data packets and secondary data packets;

[0056] The set of lost packets L i :Sink node R i The set of unwanted source packets that are not received, i.e., the sink node R i The set of sub-packets that were not received;

[0057] Request packet set W i :Sink node R i The set of source data packets that you want to receive but are lost, i.e., the destination node R i There is no set of primary packets received;

[0058] The set of unknown packets U i , Sink Node R i The set of source packets for which feedback information is lost.

[0059] S3. Each sink node sends a feedback data packet to the source node to feedback the reception status of its own source data packet. The source node constructs a state feedback matrix according to the reception status of the source data packet fed back by each sink node.

[0060] Specifically, the state feedback matrix SFM (State Feedback Matrix) is defined to represent the source data packet reception status of each sink node. In the present invention, the elements in the feedback matrix are The value range of is as follows:

[0061]

[0062] Among them, P j represents the jth source data packet, f i,j Represents the sink node R i For the receiving state of the jth source data packet, if P j Belongs to the received data packet set H i , then the element f in the feedback matrix i,j The value is 0. The set of unknown packets U i Included in the lost data packet set L i , if P j Belongs to the lost data packet set L i , but does not belong to the unknown state packet set U i, then the element f in the feedback matrix i,j The value is -1; the state of the unknown data packet set U i Contained in the request packet set W i , if P j Belongs to the request data packet set W i , and does not belong to the set of unknown packets U i , then the element f in the feedback matrix i,j The value is 1. If P j The set of packets belonging to unknown state U i , then the element f in the feedback matrix i,j The value is x.

[0063] S4. Create a confidence point set and an unconfident point set according to the request data packet set of each sink node, and use a confidence state update strategy and a likelihood estimation strategy to calculate the source data packet states in the confidence point set and the unconfident point set respectively.

[0064] Specifically, in each coded transmission process in the retransmission recovery phase, the feedback data packet of each sink node is updated, and a new confidence point set and unconfidence point set are constructed according to all updated feedback data packets. The confidence state update strategy and likelihood estimation strategy are used to calculate the source data packet states in the new confidence point set and unconfidence point set, respectively, and then the state feedback matrix is ​​updated. The specific process of step S4 is as follows:

[0065] S41. Calculate the average request set according to the size of the request data packet set in all sink node feedback data packets, place the sink nodes corresponding to the request data packet set larger than the average request set into the trusted point set, and place the sink nodes corresponding to the request data packet set not larger than the average request set into the untrusted point set;

[0066] S42. The source data packet reception conditions of all sink nodes in the confidence point set are arranged and combined, and the confidence corresponding to all combinations is calculated using a confidence formula (each combination corresponds to a confidence, which can also be understood as a transition probability). The confidence formula is expressed as:

[0067]

[0068] Wherein, b(s') represents the confidence level when the current time slot state is s', τ(b,a,z) represents the probability of reaching state z after sending coded packet a in the state corresponding to the maximum confidence b(s) of the previous time slot; b(s) represents the confidence level when the previous time slot state is s, O(s',a,z) represents the probability that the source node broadcasts coded packet a to all sink nodes, and after the system is transferred to state s', the sink node observes that the coded packet receiving state is z; T(s,a,s') represents the probability that the system is transferred to state s' after the source node broadcasts coded packet a to all sink nodes in state s; S represents the system state set;

[0069] Specifically, assuming that there are sink nodes 1 and sink node 2 in the trust point set constructed during a certain coding transmission process, for the coded packets sent in this coding transmission, both sink node 1 and sink node 2 have two situations of successful reception and failed reception. Four combinations can be obtained by permuting and combining the source data packet reception status of the sink nodes: both sink node 1 and sink node 2 receive successfully; sink node 1 receives successfully but sink node 2 fails to receive; sink node 1 fails to receive but sink node 2 receives successfully; and both sink node 1 and sink node 2 fail to receive.

[0070] S43. Select the combination corresponding to the maximum confidence, and use the source data packet reception status as the current source data packet reception status of each sink node in the confidence point set, and update the state feedback matrix;

[0071] S44. Use a likelihood estimation strategy to calculate the source data packet reception status of the destination node in the non-confident point set, and update the state feedback matrix.

[0072] Specifically, the likelihood estimation strategies include:

[0073] Assume that the source node receives the last i After receiving the feedback, try to change R i As the target receiving end, it transmits the source data packet P j The number of times the coded packet is transmitted is represented by θ ij ;

[0074] If in this θ ij If forward link data loss occurs in both coded transmissions, the receiving end R i There will be no source data packet P j The probability of the event occurring is expressed as And all uncertain elements in the feedback matrix are replaced by elements 1 or -1;

[0075] If in this θ ijIn the coded transmission, there is one or more coded transmissions sending coded packets that are received by the receiving end R i Successful reception, but all feedback information is lost, that is, both forward link and reverse link data are lost at the same time, the receiving end R i There will be no source data packet P j The feedback information of the event is expressed as (k in the formula represents θ ij The number of forward link erasure events that occurred during the coded transmission, represents the number of combinations, that is, from θ ij The number of combinations selected k times in the coded transmission, Indicates that θ occurs ij -The probability of k backward link loss, i.e. θ ij -k times the probability of feedback packet loss), and the uncertain elements in the feedback matrix are replaced by element 0.

[0076] In summary, the receiving end R i In fact, the uncertain data packet P has not been successfully received j Events The probability can be expressed as: Receiver R i In fact, the uncertain data packet P has been successfully received j Events The probability can be expressed as: Count events separately and events The probability value of the event is selected as the actual reception status of the data packet.

[0077] Specifically, during each encoding transmission process, if the receiving end successfully receives an immediately decodable packet, it needs to send feedback information, that is, a feedback data packet, to the sending end; but if the status of a source data packet is unknown, it means that the source data packet may have forward link data loss, or forward link and reverse link data loss exist at the same time; that is, all feedback data packets about the source data packet have been lost.

[0078] S5. Update the state feedback matrix, establish an immediately solvable network coding graph and calculate the weight value of each vertex.

[0079] Specifically, the vertex set of the immediately decomposable network coding graph can be divided into a main graph and a secondary graph; the main graph and the secondary graph are constructed based on the updated state feedback matrix, wherein a main data packet in the request data packet set of a destination node constitutes a vertex, and the main graph is composed of the main data packets in the request data packet set of all destination nodes; a secondary data packet in the lost data packet set of a destination node constitutes a vertex, and the secondary graph is composed of the secondary data packets in the lost data packet set of all destination nodes; the vertices in the main graph and the secondary graph are connected according to the generated edge conditions to generate the immediately decomposable network coding graph.

[0080] Specifically, assuming that there is a ij and ν kl Two vertices, vertex ν ij Represents the sink node R i The jth source packet among N source packets is lost, and vertex ν ij and ν kl If any of the following edge generation conditions are met, an edge is generated between two vertices to connect them. The edge generation conditions include:

[0081] (1) j = l, i.e., the sink node R i and the sink node R k Request retransmission of the same source packet;

[0082] (2)j∈H k And l∈H i , that is, the sink node R i and the sink node R k The source data packets requested for retransmission belong to each other's received data packet set.

[0083] In one embodiment, assuming that a source node sends 4 source data packets to 3 sink nodes, the state feedback matrix at a certain moment is as shown in Table 1:

[0084] Table 1 State feedback matrix

[0085]

[0086] It can be seen from Table 1 that the source data packets P1 and P3 are in the received data packet set H1 of the sink node R1, the source data packet P2 belongs to the request data packet set W1 of the sink node R1, and does not belong to the state unknown data packet set U1; the source data packets P1 and P2 are in the received data packet set H2 of the sink node R2, the source data packet P3 belongs to the request data packet set W2 of the sink node R1, and does not belong to the state unknown data packet set U2; the source data packet P4 is in the received data packet set H3 of the sink node R3, the source data packet P3 belongs to the request data packet set W3 of the sink node R3, and does not belong to the state unknown data packet set U3.

[0087] According to the estimated state feedback matrix and the generated edge conditions, the network coding graph can be solved, such as Figure 3 As shown, vertex V12 indicates that the destination node R1 wants to request retransmission of the lost source data packet P2; vertex V23 indicates that the destination node R2 wants to request retransmission of the lost source data packet P3; and the source data packet P2 is in the received data packet set H2 of the destination node R2, and the source data packet P3 is in the received data packet set H1 of the destination node R1, so one of the edge generation conditions is met, and a connecting edge can be generated between vertex V12 and vertex V23 to connect them to each other, and the relationship between the remaining vertices is similar.

[0088] Specifically, in each coded transmission in the retransmission recovery phase (each time an optimal retransmission coded packet is determined for broadcast), in order to reduce the completion delay of subsequent coded transmissions, the sink node that currently loses the most source data packets needs to be used as the target receiving node, and the minimum completion delay of each sink node is defined as:

[0089]

[0090] Among them, i Represents the sink node R i The minimum completion delay, W i Represents the sink node R i The request packet set, p i Represents the sink node R i The probability of source data packet loss in the forward link is q, and the transmission link from the source node to the sink node is called the forward link, and the transmission link from the sink node to the source node is called the backward link. i Represents the sink node R i The probability of source data packet loss in the backward link of , the source data packet loss rate of the previous link and the backward link both obey the Bernoulli distribution; γ represents a variable with an increase multiple, which has no effect on the result;

[0091] According to the minimum completion delay and the vertex connection relationship in the immediately solvable network coding graph, the completion delay minimization problem is transformed into a vertex weight problem, and the vertex ν is defined ij The weight is:

[0092]

[0093]

[0094] Among them, a i,k represents vertex ν ij and vertex ν kl The connection relationship, ω ij represents vertex ν ij The weight of .

[0095] S6. The source node uses the maximum weight clique search algorithm based on the weight values ​​of the vertices to obtain a coding packet and broadcasts it to all destination nodes; each destination node determines whether to update the feedback data packet and send it to the source node based on the type of the coding packet.

[0096] Specifically, the source node uses the maximum weight clique search algorithm based on the weight values ​​of the vertices to obtain the best retransmission coding packet, including:

[0097] S61. Initialize an empty coding group, determine the vertex corresponding to the maximum weight value in the immediately solvable network coding graph, and add it to the coding group;

[0098] S62. In the main graph, find a vertex that is connected to all vertices in the current coding group and has the largest weight value and add it to the coding group, and repeat this process until there is no such vertex in the main graph;

[0099] S63. In the secondary graph, find a vertex that is connected to all vertices in the current coding group and has the largest weight value and add it to the coding group, and repeat this process until there is no such vertex in the secondary graph;

[0100] S64. Finally, the data packets represented by the vertices selected by the coding group are subjected to XOR coding operation to generate the best retransmission coding packet.

[0101] Specifically, in each coded transmission in the retransmission recovery phase, the source node broadcasts the coded packets generated this time to all the sink nodes, and the coded packets can be divided into three types: non-regenerated, immediately decomposable, and non-immediately decomposable. i Any source data packet in the request data packet set, then the coded packet is for the sink node R i is a non-regenerated coding packet; if the coding packet only contains the destination node R i Any source data packet in the request data packet set, then the coded packet is for the sink node R i is an immediately decodable data packet; if the coded packet contains the destination node R i Any two or more source data packets in the request data packet set, then the coded packet is for the sink node R i This is a non-immediately decodable data packet.

[0102] In this embodiment, the sink node that receives the immediately decodable coded packet is called the target receiving end. After receiving the coded packet, the target receiving end updates the feedback data packet and sends it to the source node. The sink node that is not the target receiving end will not update the feedback data packet and send it to the source node; at the same time, due to the instability of the transmission link, the feedback data packet may be lost, which will lead to the inability to accurately update the state feedback matrix and affect the generation of the coded packet in the next encoding transmission process. Therefore, the source node first updates the state feedback matrix according to the received feedback data packet, and then re-executes step S4.

[0103] S7. Determine whether all destination nodes have completed the retransmission of the source data packet. If so, end the process; if not, return to step S4.

[0104] Specifically, there are multiple encoding transmission processes in the retransmission recovery phase. In each encoding transmission process, only one encoding packet is generated for retransmission. This encoding packet is composed of one or more source data packets encoded together. In each encoding transmission process, some destination nodes do not receive the source data packets they request to retransmit, so multiple encoding transmissions are required to ensure that all destination nodes receive the source data packets they want.

[0105] In the present invention, unless otherwise clearly stipulated and limited, the terms such as "installation", "setting", "connection", "fixation" and "rotation" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral one; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. Unless otherwise clearly defined, ordinary technicians in this field can understand the specific meanings of the above terms in the present invention according to the specific circumstances.

[0106] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A retransmission method based on immediately decomposable network coding under imperfect feedback, characterized in that: The following steps are involved: S1. Using POMDP theory to construct an imperfect feedback environment system, the imperfect feedback environment system includes a network transmission model and a POMDP model, the network transmission model includes a source node and multiple sink nodes; The POMDP theory is used to construct the POMDP model, which is defined as a six-tuple including system state, action, observation state, system state transition probability, system observation state transition probability and system immediate reward function, where: The system state set S = {s1, s2, s3, s4...} represents a set of implicit finite states of a system, and the element s i represents the i-th system state of all sink nodes when receiving coded packets; Action set A = element a in {a1,a2,a3,a4...} i Indicates the i-th type of coded packet that the source node can send in the current state; Observation state set Z = {z i ,z2,z3,z4...} i It indicates the reception status of the i-th source data packet observed by the source node based on the feedback information of all sink nodes; The system state transition probability T(s,a,s') represents the probability that the system will transition to state s' after the source node broadcasts the coded packet a to all sink nodes when the system is in state s. The system observed state transition probability O(s',a,z) represents the probability that the source node broadcasts the coded packet a to all the sink nodes, and after the system transitions to state s', the sink node observes the probability that the coded packet receiving state is z; The system immediate reward function V(s,a,s',z) represents the reward that can be obtained when the system state changes to s' when action a is taken in the system state s and the current observed state is z; S2. The source node broadcasts N source data packets to all sink nodes, and each sink node constructs a feedback data packet for the reception of the source data packet; the feedback data packet includes a set of received data packets, a set of lost data packets, a set of request data packets, and a set of unknown state data packets; S3. Each sink node sends a feedback packet to the source node to feedback its own source packet reception situation, and the source node constructs a state feedback matrix based on the source packet reception situation of each sink node feedback; S4. Create a set of trusted points and a set of untrusted points according to the request data packet set of each sink node, calculate the source data packet state in the trusted point set using a trusted state update strategy, and calculate the source data packet state in the untrusted point set using a likelihood estimation strategy; The process of step S4 includes: S41. Calculate the average request set according to the size of the request data packet set in all sink node feedback data packets, place the sink nodes corresponding to the request data packet set larger than the average request set into the trusted point set, and place the sink nodes corresponding to the request data packet set not larger than the average request set into the untrusted point set; S42. The source data packet reception conditions of all sink nodes in the confidence point set are arranged and combined, and the confidence level corresponding to each combination is calculated using a confidence formula. The confidence formula is expressed as: Wherein, b(s') represents the confidence level when the current time slot state is s', τ(b,a,z) represents the probability of reaching state z after sending coded packet a in the state corresponding to the maximum confidence b(s) of the previous time slot; b(s) represents the confidence level when the previous time slot state is s, O(s',a,z) represents the probability that the source node broadcasts coded packet a to all sink nodes, and after the system is transferred to state s', the sink node observes that the coded packet receiving state is z; T(s,a,s') represents the probability that the system is transferred to state s' after the source node broadcasts coded packet a to all sink nodes in state s; S represents the system state set; S43. Select the combination corresponding to the maximum confidence, and use the source data packet reception status as the current source data packet reception status of each sink node in the confidence point set, and update the state feedback matrix; S44. Using a likelihood estimation strategy to calculate the source data packet reception status of the destination node in the non-confident point set, and updating the state feedback matrix; S5. Update the state feedback matrix, establish an immediately solvable network coding graph and calculate the weight value of each vertex; S6. The source node uses the maximum weight clique search algorithm based on the weight values ​​of the vertices to obtain a coded packet and broadcasts it to all sink nodes; the sink node updates the feedback data packet based on the coded packet and sends it to the source node; S7. Determine whether all destination nodes have completed the retransmission of the source data packet. If so, end the process; if not, return to step S4.

2. The retransmission method based on immediately decomposable network coding under imperfect feedback according to claim 1, characterized in that: The network transmission model includes a source node and multiple sink nodes. The source node broadcasts N source data packets to all sink nodes through a sending channel. Each sink node receives all or part of the source data packets according to its own needs, and sends feedback data packets through a feedback channel. Both the sending channel and the feedback channel are deletion channels.

3. The retransmission method based on immediately decomposable network coding under imperfect feedback according to claim 1, characterized in that: The feedback data packet of each sink node includes four data packet sets: Received data packet set H i :Sink node R i The set of successfully received source data packets, including primary data packets and secondary data packets; The set of lost packets L i :Sink node R i The set of unwanted source packets that were not received; Request packet set W i :Sink node R i The set of source packets that were intended to be received but were lost; The set of unknown packets U i , Sink Node R i The set of source packets for which feedback information is lost.

4. The retransmission method based on immediately decomposable network coding under imperfect feedback according to claim 1 or 3, characterized in that: In step S5, a main graph and a secondary graph are constructed based on the updated state feedback matrix, wherein the main graph is formed by source data packets in the request data packet set of all destination nodes; and the secondary graph is formed by source data packets in the non-request data packet set in the lost data packet set of all destination nodes; both the main graph and the secondary graph are vertex sets; the vertices in the main graph and the secondary graph are connected according to the edge generation conditions to generate an immediately decomposable network coding graph.

5. The retransmission method based on immediately decomposable network coding under imperfect feedback according to claim 4, characterized in that: Assume that there is ij and ν kl Two vertices, vertex ν ij Represents the sink node R i The jth source packet among N source packets is lost, and vertex ν ij and ν kl If any of the following edge generation conditions are met, an edge is generated between two vertices to connect them. The edge generation conditions include: 1.j=l, i.e., the destination node R i and the sink node R k Request retransmission of the same source packet; 2.j∈H k And l∈H i , that is, the sink node R i and the sink node R k The source data packets requested for retransmission belong to each other's received data packet set.

6. The retransmission method based on immediately decomposable network coding under imperfect feedback according to claim 1, characterized in that: The formula for calculating the weight value of each vertex in the immediately solvable network coding graph is: Among them, a i,k represents vertex ν ij and vertex ν kl The connection relationship, ω ij represents vertex ν ij The weight of i Represents the sink node R i The minimum completion delay, W i Represents the sink node R i The request packet set, p i Represents the sink node R i The probability of source data packet loss in the forward link is called the forward link; γ represents a variable with an increase multiple.

Images