A reliable data transmission method for underwater acoustic network based on online fountain code

Abstract

The present invention provides a reliable transmission method for underwater acoustic network data based on online fountain codes. The dynamic parity sequence coding method proposed in the establishment stage realizes the expansion of basic components through rule coding, reduces the number of interconnected components, and thus reduces the number of coding packets and the number of feedbacks required in the completion stage. In the completion stage, the sending of invalid coding packets is avoided as much as possible, and the receiving party's interconnected components are determined by updating the receiving party's decoding diagram, and representative data packets are screened from each component and fed back to the sender. The present invention effectively reduces the number of invalid coding packets sent in the completion stage, thereby reducing the decoding overhead in the completion stage; it also reduces the feedback overhead of the online fountain code, and can realize reliable transmission of data packets with low overhead in underwater acoustic networks with limited bandwidth and energy consumption, reducing the overall decoding overhead, improving the efficiency of data packet transmission in the underwater acoustic network, improving the decoding efficiency, and reducing the overall overhead.

Description

Technical Field

[0001] The present invention relates to the field of underwater communication networking, and specifically to underwater acoustic communication networks, wireless communication networks, data link protocols, forward error correction technologies, etc. Specifically, it is a reliable transmission method for underwater acoustic communication networks. The method can be used for underwater acoustic communication networks and wireless communication networks. Background Art

[0002] Underwater Acoustic Sensor Networks (UASNs) are communication networks covering three-dimensional underwater areas, which are connected by acoustic links and are composed of a series of communication nodes distributed on the surface and underwater. With the development of science and technology, UASNs have been widely used in both military and civil fields. As human demand for resources intensifies, the status of underwater acoustic sensor networks is becoming increasingly important. Point-to-point transmission in UASNs mainly solves the reliability problem of direct mutual transmission between nodes, involving the study of reliable transmission methods for underwater data. However, in the complex and changeable marine environment, due to the absorption of water, UASNs use acoustic wave communication with low loss and long propagation distance; the propagation speed of acoustic waves is 1500m / s, which causes large propagation delays in UASNs; in addition, the bandwidth of the underwater acoustic channel is narrow, the bit error rate is high, and the noise interference of the underwater acoustic channel is large. These characteristics of the underwater environment make it easy for data transmission between underwater sensor nodes to be erroneous or even lost. Therefore, in order to ensure the reliable transmission of data, higher requirements will inevitably be placed on the reliable transmission method in UASNs.

[0003] Packet-Level FEC erasure codes use the redundancy of data packets to support reliable data transmission over lossy paths. This scheme encodes the original message of k data packets into a set of coded messages containing n coded packets so that the original message can be recovered from a subset of the coded messages. The coded packets are sent to the receiver as redundancy, and the coded packets can be sent until the k original data packets are fully recovered. Among them, digital fountain codes with rateless characteristics are suitable for highly dynamic UASNs channels, because the number of coded packets as redundancy can be dynamically adjusted, so that the data can adapt to different error rates during the propagation process until full recovery is achieved. This overcomes the difficulty of traditional packet-level FEC schemes in code rate compatibility when the channel changes continuously. Luby transform (LT) code is the first practical fountain code. However, when fountain codes are used as a reliable data transmission scheme in UASNs, although its rateless characteristics can well adapt to the continuous changes of the channel over time, its large decoding overhead is not friendly to network nodes in UASNs. Online Fountain Codes (OFC) retain the rateless characteristics of fountain codes and utilize the idea of ​​feedback channels. The receiver feeds back the decoding status, and the sender selects the optimal encoding strategy based on the feedback information, thereby improving the useful probability of the coded packet and reducing the total number of coded packets required for complete decoding. OFC further simplifies the encoding and decoding process of LT, eliminates the error flattening by feedback, and obtains good partial decoding capability and real-time performance; most importantly, compared with the scheme using fountain codes, it reduces the decoding overhead and the number of feedbacks. At present, many scholars have studied the data transmission method based on online fountain codes. Among them, the improved online fountain code (Improved Online Fountain Code, IOFC) proposes a relationship framework between the overhead and the number of recovered original packets based on random graph theory, and introduces a method for selecting the specified original packet in the establishment phase. IOFC has lower overhead than OFC, but does not solve the problem of the decrease in the useful probability of the coded packet in the completion phase. In order to improve the decoding recovery rate in the intermediate process of OFC, the online fountain code without the establishment phase (Online Fountain Codes without Build-up phase, OFCNB) first transmits the coded packets of degree 1 and removes the establishment phase to improve the intermediate decoding performance, but there is still a problem of high decoding overhead. Aiming at the problem that the useful probability of coded packets gradually decreases in the completion stage of OFC coding, a recursive online fountain code (ROFC) is proposed in combination with the application of OFC in UASNs. The recursive coding strategy feeds back the set of successfully decoded data packet sequences, so that the sender can obtain the undecoded data packet sequence.Selecting the original data packets involved in encoding based on the sequence of undecoded data packets greatly improves the probability that the encoded packets can advance decoding in the completion phase, but there is still a problem of high decoding overhead in the establishment phase. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention provides a reliable transmission method for underwater acoustic network data based on online fountain codes. In view of the problem of large decoding overhead when online fountain codes (OFC) are applied to UASNs as a packet-level FEC scheme for encoding and transmitting, the present invention proposes an online fountain code transmission method based on dynamic parity sequence and decoding graph update feedback. The sending end adopts a dynamic parity sequence encoding strategy, and the receiving end updates the Unicom component based on the decoding graph to feedback the candidate encoding data packet set, aiming to significantly reduce the encoding and decoding overhead of the online fountain code in the underwater acoustic network transmission process, while improving the reliability of underwater acoustic network data transmission, effectively improving its efficiency and applicability in the underwater acoustic network.

[0005] The overall process of the online fountain code transmission method based on dynamic parity sequence and decoding graph update feedback is divided into the establishment stage and the completion stage. The dynamic parity sequence encoding method proposed in the establishment stage realizes the expansion of basic components through rule encoding, reduces the number of interconnected components, and thus reduces the number of coding packets required in the completion stage and the number of feedbacks. In the completion stage, in order to ensure that the transmitted coding packets are useful for decoding, that is, to avoid the transmission of invalid coding packets as much as possible, the receiving party's interconnected components are determined by updating the receiving party's decoding graph, and representative data packets are screened from each component and fed back to the sender. This feedback mechanism based on decoding graph update effectively reduces the number of invalid coding packets sent in the completion stage, thereby reducing the decoding overhead in the completion stage; on this basis, the present invention also reduces the feedback overhead of the online fountain code by setting the feedback threshold. Based on the above design, the online fountain code transmission method based on dynamic parity sequence and decoding graph update feedback proposed in the present invention can realize reliable transmission of data packets with low overhead in an underwater acoustic network with limited bandwidth and energy consumption.

[0006] The detailed steps of the technical solution adopted by the present invention to solve the technical problem are as follows:

[0007] Step 1: Rule encoding in the establishment phase;

[0008] The sender needs to transmit block data containing k original packets. In the initial stage, the decoding graph (unipartite graph) maintained by the receiver has k vertices and no edges. Encoding and decoding exist in both the establishment and completion stages. The establishment stage aims to connect and decode the original data packets of a preset size of kβ0, where β0 represents the ratio of the size of the largest connected component to the total number of original data packets k. The completion stage aims to complete the decoding of components other than the largest connected component.

[0009] In the establishment phase, receive the coded packets of degree 2, that is, randomly increase the edges between the vertices in the decoding graph at the receiving end until the maximum number of connected components reaches kβ0. The decoding graph is g(k,p), p is the probability of each edge being selected, p=c / k, c is the average number of times each node is selected, and a node in the decoding graph corresponds to an original data packet. There are k nodes in the graph g(k,p) and at most k(k-1) / 2 edges. It is known that the relationship between c and β0, that is, the proportion of the maximum connected component to the total number of data packets β0 and the average number of times each original packet is selected c, satisfies the following relationship:

[0010]

[0011] E(N B2_i ) represents the expected number of degree 2 coded packets sent for the i-th time during the establishment phase, E(N B1 ) represents the expected number of degree 1 coded packets sent during the establishment phase, E(N C1,C2 ) is the expected number and number of degree 1 and degree 2 coded packets sent in the completion phase. E(N total ) represents the expected number of total coded packets required to transmit k packets. N i (Com_Syb) represents the set of candidate packets encoded by the sender after the ith update in the completion phase;

[0012] Take L=3 as an example to describe the detailed steps of the solution:

[0013] Step 1: Label all original data packets 1, 2, 3...k; select two adjacent data packets for XOR encoding and send until all data packets have participated in the encoding; at this time, the receiver's decoding graph g(k1, p1) is updated to a random graph with k1=k / 2 vertices; the number of encoded packets sent is:

[0014]

[0015] Step 2: To reduce the number of interconnected components and increase the size of components, the original packets with even numbers are sent in pairs in XOR order until all packets with even numbers have been encoded. The receiver's decoding graph g(k2,p2) is updated to a random graph with k2=k / 4 vertices. The number of coded packets sent is:

[0016]

[0017] Step 3: To further reduce the number of interconnected components and increase the size of the components, two adjacent original data packets in the same order of sequence are XOR-encoded and sent from the original packets with odd sequence numbers that are divisible by 4 with a remainder of 1; until the data packets with the specified sequence numbers have participated in the encoding; the receiver's decoding graph g(k3,p3) is updated to a random graph with k3=k / 8 vertices; the number of coded packets sent is:

[0018]

[0019] Step 2: Random coding during the establishment phase;

[0020] The preset maximum interconnected component size ratio is β0. After completing the rule encoding in step 1, the sender randomly selects two original data packets from all original data packets and generates a coded packet of degree 2 using XOR encoding and sends it. After receiving the coded packet, the receiver updates the decoding graph and determines the maximum interconnected component size in the decoding graph. The sender continues to send degree 2 coded packets and update the decoding graph until the maximum interconnected component size of the receiver has reached kβ0, and feedback is given to the sender containing the information of "sending a coded packet of degree 1". At this time, the decoding graph is g(k3,p3)=g(k3,c / k3), where k3=k / 8. The expected number of degree 2 coded packets required to construct a interconnected component of size kβ0 is:

[0021]

[0022] Step 3: Decoding of the maximum connected component is performed during the establishment phase;

[0023] After the sender receives the feedback packet containing the information of "sending the coded packet of degree 1", it proves that the constructed maximum connected component needs to be decoded; the sender randomly selects an original packet from all the original data packets and sends it as the coded packet of degree 1. After receiving the coded packet, the receiver decodes it and updates the decoding graph; repeat this process until the receiver successfully decodes the maximum connected component and feeds back a feedback packet containing the information of "maximum connected component decoding success + encoding candidate original data packet set" to the sender; the expected number of sent degree 1 coded packets is:

[0024]

[0025] Step 4: The completion phase performs decoding of the remaining components;

[0026] The sender will decode the feedback packet information successfully based on the largest component according to the encoding candidate original data packet set N i(Com_Syb) instructs the sender to generate and send a coded packet; selects original data packets from the candidate set according to the probabilities of the degree 1 and degree 2 coded packets, generates and sends degree 1 and degree 2 coded packets; if the receiver receives a degree 2 coded packet, the receiver updates the decoding map and the interconnection components; if a degree 1 coded packet is received, it decodes it and selects a candidate coding set based on the undecoded interconnection components, and feeds back a feedback packet containing the "coding candidate original data packet set" information to the sender, and the sender updates the set N i (Com_Syb), the sender continues to send degree 1 coded packets and degree 2 coded packets, and the receiver continues to decode according to the received coded packets until all the original data packets are decoded successfully.

[0027] In step 1, the sender performs sequential rule encoding, encodes and sends a coded packet of degree 2 according to the preset sequence number rule, that is, obtained by XOR encoding two original data packets. The rule encoding stage is divided into L steps, L is constrained by the number of original data packets k and the preset maximum Unicom component size β0, that is, L <log2(kβ0)。

[0028] In step 4, the expected number of coded packets sent is:

[0029]

[0030] The sum of the probabilities of the preset degree 1 and degree 2 coded packets is 1, the probability of sending a degree 1 coded packet is p1, and the probability of a degree 2 coded packet is p2 = 1-p1; after the end of step 3, after removing the largest connected component from the random graph g = (k, c / k), the remaining nodes and edges also form a random graph g' = (t, d / t); wherein t = (1-β0)k, d = (1-β0)c, t represents the number of original packets that have not been decoded, d represents the average number of times the original packets in the graph g' are selected, that is, the average degree of the original packets; in the graph g' = (t, d / t), before the completion stage, t·d represents the total number of times t original packets are selected, and td / 2 edges are generated; since the sent degree 1 and degree 2 coded packets are both useful for decoding, the expected number of coded packets required to be sent in the fourth step is:

[0031]

[0032] Right now:

[0033]

[0034] In summary, to restore k original packets, we can get:

[0035]

[0036] Among them, E(N total) is the number of coded packets expected to be transmitted.

[0037] In step 4, the proportion of the decoded data packets in the total data packets is calculated. If the difference in the proportions exceeds the preset threshold, that is, β(j)-β(j-1)<δ, a feedback packet containing the information of the "encoding candidate original data packet set" is fed back to the sender, and the updated set N is sent. i (Com_Syb), continue to generate and send coded packets of degree 1 and degree 2, and repeat the process until all original data packets are decoded successfully; where: β(j) is the ratio of the original data packets that have been successfully decoded at the jth decoding to the total original data packets, and δ is the preset feedback threshold value.

[0038] The beneficial effects of the present invention are:

[0039] 1. In the coding establishment phase, firstly, the coding is performed through L layers (e.g. L=3) of regular coding, and the decoding graph g(k,c) originally containing k vertices is transformed into g(k / 8,c / 8). This means that in theory all the original data packets are involved in the coding, which greatly reduces the number of interconnected components. This reduces the overhead required for decoding in the completion phase, thereby reducing the overall decoding overhead and improving the data packet transmission efficiency in the underwater acoustic network. In addition, the retained rateless characteristics in the random coding phase can adapt to changes in channel conditions.

[0040] 2. In the completion phase of encoding, the decoding graph and the undecoded original data packets are integrated at the receiving end. All undecoded connected components are selected in the decoding graph, and one data packet is randomly selected from each component to be included in the feedback set. Single data packets that are not connected due to channel packet loss are directly included in the feedback set. The coded packets generated by the sender after random selection can connect two components or decode a certain component, which reduces the probability of useless coded packets, reduces the decoding flat layer problem in the completion phase, improves decoding efficiency, and reduces overall overhead.

[0041] The above characteristics endow the method of the present invention with good performance, and when there is channel packet loss, it can also achieve a lower overall overhead. After simulation experiments, it is verified that the method of the present invention has significantly reduced decoding overhead and feedback overhead compared with several existing methods, thereby improving the overall transmission performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is an overall flow chart of the reliable transmission method of water acoustic network data based on online fountain code of the present invention.

[0043] Figure 2It is a schematic diagram of the rule coding scheme in the establishment stage of this method. Among them, (a), (c), and (e) are the flow schematic diagrams of the present invention in the rule coding stage. (a), (b), and (d) are the flow schematic diagrams of the rule coding stage for comparison.

[0044] Figure 3 It is a schematic diagram of selecting the coding candidate data packet set in the completion stage of the present invention.

[0045] Figure 4 It is a comparison chart of the scheme performance based on the method of the present invention. Figure 4 (a) is a comparison chart of the decoding overhead for k = 100 - 1000. Figure 4 (b) is a comparison chart of the decoding overhead for k = 100 - 1000. Detailed implementation manners

[0046] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0047] Figure 1 It is the overall flowchart of the present invention. More specifically, the overall process is divided into an establishment (Build_Up) stage (Steps 1 - 3) and a completion stage (Step 4). In the establishment stage, coded data packets need to be sent so that a connected component of a predetermined size can be constructed at the receiving end; then the connected component is decoded, and after successful decoding, the establishment stage ends and the completion stage is entered. In the completion stage, it is determined whether to update the candidate original data packet set for the next encoding through feedback according to the proportion of the original data packets successfully decoded each time. The sender encodes according to the candidate set until all data packets are successfully decoded and the completion stage ends.

[0048] Step 1: Rule coding (establishment stage)

[0049] Determine the number of layers L of rule coding, where L < log2(kβ0). In rule coding, in the m-th (1 ≤ m ≤ L) step, when m is odd, the serial numbers of the original data packets for coding candidates are {1, 1 + 2 m-1 , 1 + 2·2 m-1 ,...}; when m is even, the serial numbers of the original data packets for coding candidates are {2, 2 + 2 m-1 , 2 + 2·2 m-1 ,...}. The following takes L = 3 to describe the specific implementation manner of the scheme.

[0050] The first step: All original packets are arranged in serial number order. Starting from the 1st original packet, select two consecutive original packets for exclusive OR coding to generate a coding packet of degree 2. Then one original packet is marked as D i , and the serial number of this original packet is i; the other original packet is marked as D i+1 , with the serial number i + 1; the coding packet is marked as Cj , the serial number is j; then If k is an odd number, then i=1,3,5,...k-2, If k is an even number, then i = 1, 3, 5, 7 ... k - 1. The sender sends a coded packet of degree 2 to the receiver until the encoding of this step is completed.

[0051] Step 2: In this step, all the original packets with even numbers are encoded in pairs according to the sequence number to generate a coded packet of degree 2, that is, C j =D i+2 ⊕D i ,in The sender sends a degree 2 coded packet to the receiver until the encoding of this step is completed. Taking into account the packet loss of the channel, if the adjacent data packets with cardinality numbers are selected for encoding and sending, the original data packets that failed to connect successfully in the first step will not be selected in this step and become separate components. If the coded packet of this step is successfully received, selecting the adjacent original data packets with even numbers will overcome the failed connection in the first step, thereby reducing the decoding level that occurs in the completion stage and reducing the overall transmission overhead. The encoding diagram is shown as follows Figure 3 shown.

[0052] Step 3: From the original packets with odd numbers that are divisible by 4 with a remainder of 1, encode them two by two in the same order of numbers to generate a coded packet of degree 2, that is, in The sender sends the coded packet of degree 2 to the receiver until the encoding of this step is completed, and the encoding is completed according to the sequence number rule.

[0053] Step 2: Random Encoding (Establishment Phase)

[0054] The preset maximum connected component size ratio is β0. After the rule encoding, the sender randomly selects two original data packets from all the original packets, XORs them to generate a coded packet of degree 2 and sends it. After receiving the coded packet, the receiver updates the decoding graph and determines the maximum connected component size in the decoding graph. Repeat this process until the maximum connected component size of the receiver has reached kβ0, and feedback is given to the sender with a feedback packet containing the information of "sending a coded packet of degree 1".

[0055] Step 3: Decoding of the largest connected component (establishment phase)

[0056] After the sender receives the feedback packet containing the information "sending the coded packet of degree 1", it proves that the constructed maximum connected component needs to be decoded. The sender randomly selects an original packet from all the original packets and sends it as the coded packet of degree 1. After receiving the coded packet, the receiver decodes it and updates the decoding graph. Repeat this process until the receiver successfully decodes the maximum connected component and feeds back a feedback packet containing the information "successful decoding of the maximum connected component + set of coded candidate original data packets" to the sender.

[0057] Step 4: Decoding of remaining components (completion phase)

[0058] The sender will decode the feedback packet information successfully based on the largest component according to the encoding candidate original data packet set N i (Com_Syb) is used to guide the sender to generate and send coded packets. According to the respective probabilities of degree 1 and degree 2 coded packets, the original data packets are selected from the candidate set, and coded packets of degree 1 and degree 2 are generated and sent. If the receiver receives a degree 2 coded packet, the receiver updates the decoding diagram and updates the interconnection component; if a degree 1 coded packet is received, it is decoded and a candidate coding set is selected based on the undecoded interconnection components. Taking into account the long delay characteristics of the underwater acoustic channel in practice, in order to improve the applicability of the data transmission method in UASNs, the feedback overhead is reduced by limiting the number of feedbacks. That is, by calculating the proportion of the decoded data packet in the total data packet, if it exceeds the preset threshold, that is, β(j)-β(j-1)<δ, a feedback packet containing the "coding candidate original data packet set" information is fed back to the sender, and the updated set N is sent. i (Com_Syb), continue to generate and send coded packets of degree 1 and degree 2, and repeat the process until all original packets are decoded successfully. Where: β(j) is the proportion of the original data packets that have been successfully decoded in the jth decoding to the total original data packets. δ is the preset feedback threshold value.

[0059] Figure 3 The black color indicates the original data packets that have been successfully decoded, and the white color indicates the data packets that have not been decoded. When selecting the candidate original data packets for coding, a data packet is randomly selected from each undecoded Unicom component in the decoding diagram of the receiver. When there is only one data packet in the component, it is directly included in the set. Compared with the feedback of all undecoded original data packets in ROFC-LF, the selection strategy in the present invention ensures that even if the size of the undecoded Unicom component is large, the sent degree 2 coded packet will not become a useless coded packet, that is, the useless probability of the sent coded packet is reduced. The decoding flat layer problem that occurs in the completion stage is reduced, so as to achieve the purpose of improving decoding efficiency and reducing overall overhead.

[0060] Figure 4In , DPOFC is the method of the present invention, where L = 3. OFC, IOFC, and ROFC are existing methods. The number of original data packets ranges from 100 to 1000, the maximum Unicom component size is set to 0.5, the feedback limit threshold is set to 0.006, and the channel packet loss rate is 0.1. Figure 4 It can be seen that the overall overhead of the present invention is significantly less than that of OFC, IOFC and ROFC-LF solutions. That is, the performance is improved in terms of decoding overhead and feedback overhead. This is enough to show the superiority of the transmission method described in the present invention.

Claims

1. A reliable transmission method of water acoustic network data based on online fountain code, characterized in that The following steps are included: Step 1: Rule encoding is performed in the establishment phase; The sender needs to transmit a block of k original packets. The decoding graph maintained by the receiver in the initial stage has k vertices and no edges. Encoding and decoding exist in both the establishment and completion stages. The establishment stage aims to connect and decode the original data packets of a preset size of kβ0, where β0 represents the ratio of the size of the largest connected component to the total number of original data packets k. The completion stage aims to complete the decoding of components other than the largest connected component. In the establishment phase, receive a coded packet of degree 2, that is, randomly increase the edges between the vertices in the decoding graph at the receiving end until the maximum number of connected components reaches kβ0. The decoding graph is g(k,p), p is the probability of each edge being selected, p=c / k, c is the average number of times each node is selected, and a node in the decoding graph corresponds to an original data packet. There are k nodes and at most k(k-1) / 2 edges in the graph g(k,p). The relationship between c and β0, that is, the ratio of the maximum connected component to the total number of data packets β0 and the average number of times each original packet is selected c, satisfies the following relationship: E(N B2_i ) represents the expected number of degree 2 coded packets sent for the i-th time during the establishment phase, E(N B1 ) represents the expected number of degree 1 coded packets sent during the establishment phase, E(N C1,C2 ) is the expected number and number of degree 1 and degree 2 coded packets sent in the completion phase; E(N total ) represents the expected number of total coded packets required to transmit k packets; N i (Com_Syb) represents the set of candidate packets encoded by the sender after the ith update in the completion phase; Step 1: Label all original data packets 1, 2, 3...k; select two adjacent data packets for XOR encoding and send until all data packets have participated in the encoding; at this time, the receiver's decoding graph g(k1, p1) is updated to a random graph with k1=k / 2 vertices; the number of encoded packets sent is: Step 2: To reduce the number of interconnected components and increase the size of components, the original packets with even numbers are sent in pairs in XOR order until all packets with even numbers have been encoded. The receiver's decoding graph g(k2,p2) is updated to a random graph with k2=k / 4 vertices. The number of coded packets sent is: Step 3: To further reduce the number of interconnected components and increase the size of the components, two adjacent original data packets in the same order of sequence are XOR-encoded and sent from the original packets with odd sequence numbers that are divisible by 4 with a remainder of 1; until the data packets with the specified sequence numbers have participated in the encoding; the receiver's decoding graph g(k3,p3) is updated to a random graph with k3=k / 8 vertices; the number of coded packets sent is: Step 2: Random coding during the establishment phase; The preset maximum interconnected component size ratio is β0. After completing the rule encoding in step 1, the sender randomly selects two original data packets from all original data packets and generates a coded packet of degree 2 using XOR encoding and sends it. After receiving the coded packet, the receiver updates the decoding graph and determines the maximum interconnected component size in the decoding graph. The sender continues to send degree 2 coded packets and update the decoding graph until the maximum interconnected component size of the receiver has reached kβ0, and feedback is given to the sender containing the information of "sending a coded packet of degree 1". At this time, the decoding graph is g(k3,p3)=g(k3,c / k3), where k3=k / 8. The expected number of degree 2 coded packets required to construct a interconnected component of size kβ0 is: Step 3: Decoding of the maximum connected component is performed during the establishment phase; After the sender receives the feedback packet containing the information of "sending the coded packet of degree 1", it proves that the constructed maximum interconnected component needs to be decoded; the sender randomly selects an original packet from all the original data packets and sends it as the coded packet of degree 1. After receiving the coded packet, the receiver decodes it and updates the decoding map; Repeat this process until the receiver successfully decodes the largest connected component and feeds back a feedback packet containing the information "successful decoding of the largest connected component + set of encoding candidate original data packets" to the sender; the expected number of degree 1 encoded packets sent is: Step 4: The completion phase performs decoding of the remaining components; The sender will decode the feedback packet information successfully based on the largest component according to the encoding candidate original data packet set N i (Com_Syb) instructs the sender to generate and send a coded packet; selects original data packets from the candidate set according to the probabilities of the coded packets of degree 1 and degree 2, generates and sends coded packets of degree 1 and degree 2; if the receiver receives a coded packet of degree 2, the receiver updates the decoding map and the interconnection components; if a coded packet of degree 1 is received, it decodes it and selects a candidate coding set based on the interconnection components that have not been decoded, and feeds back a feedback packet containing the "coded candidate original data packet set" information to the sender, and the sender updates the set N i (Com_Syb), the sender continues to send degree 1 coded packets and degree 2 coded packets, and the receiver continues to decode according to the received coded packets until all the original data packets are decoded successfully.

2. The reliable transmission method of underwater acoustic network data based on online fountain code according to claim 1 is characterized by: In step 1, the sender performs sequential rule encoding, encodes and sends a coded packet of degree 2 according to the preset sequence number rule, that is, obtained by XOR encoding two original data packets. The rule encoding stage is divided into L steps, L is constrained by the number of original data packets k and the preset maximum Unicom component size β0, that is, L <log2(kβ0)。 3. The reliable transmission method of underwater acoustic network data based on online fountain code according to claim 1 is characterized by: In step 4, the expected number of coded packets sent is: The sum of the probabilities of the preset degree 1 and degree 2 coded packets is 1, the probability of sending a degree 1 coded packet is p1, and the probability of a degree 2 coded packet is p2 = 1-p1; after the end of step 3, after removing the largest connected component from the random graph g = (k, c / k), the remaining nodes and edges also form a random graph g' = (t, d / t); wherein t = (1-β0)k, d = (1-β0)c, t represents the number of original packets that have not been decoded, d represents the average number of times the original packets in the graph g' are selected, that is, the average degree of the original packets; in the graph g' = (t, d / t), before the completion stage, t·d represents the total number of times t original packets are selected, and td / 2 edges are generated; since the sent degree 1 and degree 2 coded packets are both useful for decoding, the expected number of coded packets required to be sent in the fourth step is: Right now: In summary, to restore k original packets, we can get: Among them, E(N total ) is the number of coded packets expected to be transmitted.

4. The reliable transmission method of underwater acoustic network data based on online fountain code according to claim 1 is characterized by: In step 4, the proportion of the decoded data packets in the total data packets is calculated. If the difference in the proportions exceeds the preset threshold, that is, β(j)-β(j-1)<δ, a feedback packet containing the "encoding candidate original data packet set" information is fed back to the sender, and the updated set N is sent. i (Com_Syb), continue to generate and send coded packets of degree 1 and degree 2, and repeat the process until all original data packets are decoded successfully; where: β(j) is the ratio of the original data packets that have been successfully decoded at the jth decoding to the total original data packets, and δ is the preset feedback threshold value.

Images