CN114900273B - A continuous reliable transmission method based on covering window and non-uniform sampling fountain code - Google Patents

Abstract

The invention relates to a continuous and reliable transmission method based on a coverage window and non-uniform sampling fountain codes, which comprises the following steps: 1) Dividing the data blocks of the transmitted data according to the time delay requirement; 2) Equalizing sampling times of each source code based on a coding strategy of non-uniform sampling, and generating codes according to a coding mode of fountain codes; 3) And (3) continuously transmitting the source data based on the codes generated in the step (2) by adopting a transmission mode of covering windows according to the division mode of the step (1). Compared with the prior art, the method has the advantages of reducing redundancy, reducing average influence of lost partial codes and the like.

Description

A continuous reliable transmission method based on covering window and non-uniform sampling fountain code

Technical Field

The present invention relates to the field of network coding and transmission technology, and in particular to a continuous reliable transmission method based on a covering window and a non-uniform sampling fountain code.

Background technique

For transmission channels, especially in wireless ad hoc network scenarios, transmission reliability is one of the important evaluation indicators. Communication systems usually use the ARQ mechanism to deal with lossy channels. In the ARQ mechanism, the sender retransmits information based on the feedback from the receiver to ensure reliable data transmission. However, when network congestion or packet loss occurs, the sender waits for the feedback timeout and retransmits the information, which wastes bandwidth resources and increases latency. To avoid such problems, another idea is to use channel coding with error correction capabilities to combat data loss, among which the typical channel error correction coding is FEC technology. In FEC-based transmission, the sender adds some redundancy to the transmitted data through channel coding. Due to packet loss, the received information at the receiver is incomplete. However, based on the received redundant information, the lost information can be restored, avoiding the waste of bandwidth resources by the ARQ mechanism. Traditional FEC codes include Turbo codes, RS codes, LDPC codes, etc., which are codes with fixed code rates. However, due to the uncertainty of the channel state, some data cannot be successfully recovered after using fixed-rate FEC encoding. Therefore, if you want to receive the transmitted data completely, you need a small number of ARQ requests to ensure its reliable transmission. Another FEC encoding idea is rateless coding, such as fountain codes. Its encoder can theoretically generate any number of codes without specifying a code rate, so it is called rateless code. The receiver attempts to decode based on the received data. If the decoding is successful, the transmission of the current source code block is terminated. According to Shannon's second theorem, when the code block is infinitely long, the throughput can reach the Shannon limit. However, in actual systems, due to delay requirements, the source code block needs to be divided into multiple short blocks and then transmitted continuously. For a single short block, the coding redundancy of the fountain code increases significantly as the code block becomes shorter. Therefore, how to improve the redundancy of short-code-length fountain codes is one of the important issues that need to be solved at present.

Summary of the invention

The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a continuous reliable transmission method based on covering window and non-uniform sampling fountain code.

The purpose of the present invention can be achieved by the following technical solutions:

A continuous reliable transmission method based on a cover window and a non-uniform sampling fountain code, the method comprising:

S1: Divide the transmitted data into data blocks according to the delay requirements.

S2: The coding strategy based on non-uniform sampling balances the sampling times of each source code and generates the code according to the encoding method of fountain code.

S3: Send source data based on the code generated by S2, and transmit continuously using a covering window transmission method according to the division method of S1.

Furthermore, the specific content of S1 is:

First, the transmitted data is divided into data blocks of the same size, and then each divided data block is independently encoded and transmitted in blocks; after each data block is transmitted, the receiving end sends an ACK message to notify the sending end to stop sending this data block; after receiving the ACK, the sending end stops sending this data block and sends the next data block instead, and repeats the above process; among them, the overlay window overlaps the previous and next data, and when encoding, the overlay part is included in the encoding as part of this data block.

Furthermore, the specific content of S2 is:

First, sample the degree value d from the degree function and compare it with the threshold d ^* . When the degree value d≤d ^* , sample directly from the source code set. Otherwise, select d source codes from the source code set in turn, and calculate the probability of adding them to the encoding set based on the number of times they are sampled. Then decide whether to add them based on this probability, and repeat this process until d source codes are selected from the source code set. Then generate the code based on the source code set according to the XOR encoding method in the fountain code.

Furthermore, in the coding strategy based on non-uniform sampling to balance the sampling times of each source code, the number of times the current source code is sampled is used to calculate the probability of whether to add sampling, and whether the source code is added to the encoding is determined based on the calculated probability. This process is then repeated until the number of source codes that meet the requirements is sampled.

Furthermore, the specific content of S3 is:

The source data is sent based on the code generated by S2, and the covering window transmission method is adopted according to the division method in S1, so that each segment of data is dependent on each other, and the data of the previous decoded block is added to the decoding as the prerequisite information of the next block of data, and the code elements after block encoding are sent using the fountain code; the receiving end only retains the block information and coding information of the current data packet, and continuously tries to decode according to the received packets. When the receiving end completes the decoding, the receiving end stops receiving the data packet of this data block, and the sending end sends the next data block to realize continuous data transmission.

Furthermore, in the transmission mode of the covering window, the source code of the previous successfully decoded block is added to the encoding of the next block as the pre-information of the decoding of this block, that is, it is added to the decoding as a node with a degree of 1 during decoding.

The continuous reliable transmission method based on cover window and non-uniform sampling fountain code provided by the present invention has at least the following beneficial effects compared with the prior art:

1) The method of the present invention improves the performance of the decoding end through a continuous transmission strategy based on a covering window; by dynamically updating the sampling probability according to the sampling number during the encoding process, the uniformity of the number of source code samplings is improved, and the redundancy of the short code and long fountain code is reduced.

2) The present invention improves the sampling strategy, so that the average number of times the source code is encoded is significantly increased, and the difference between them is smaller, and the encoding is fairer; from the decoding end, the average impact of the lost part of the encoding is small.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of block transmission based on a window transmission strategy of a covering window in an embodiment;

FIG2 is a schematic diagram of a flow chart of a coding strategy based on non-uniform sampling in an embodiment;

FIG3 is a reliable transmission protocol message design in an embodiment;

FIG4 is a design of a source code packet index message segment in an embodiment;

FIG5 is a diagram showing a network data encapsulation process in an embodiment;

FIG6 is a network protocol stack design according to an embodiment;

FIG. 7 is a design of a transceiver cache based on a fountain code reliable transmission protocol in an embodiment.

Detailed ways

The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should belong to the scope of protection of the present invention.

Example

The present invention relates to a continuous reliable transmission method based on a covering window and a non-uniform sampling fountain code, which is used to solve the problem of network transmission reliability in a multi-hop network. The specific scheme of the method is:

Step 1. The method of the present invention adopts a window transmission strategy based on a covering window. The transmitted data is divided into blocks according to the delay requirements; in a specific communication scenario, according to the delay requirements, the data needs to be divided, and after the division, it is continuously transmitted in a covering window manner. After the current block is encoded and decoded, it moves to the next data block. At this time, the coding delay is proportional to the window length, and the coding source can be set to bit data or single packet data according to different tasks.

Specifically: during transmission, the data is first divided into blocks of the same size, and then each block is encoded independently and transmitted in blocks as shown in Figure 1. After each block is transmitted, the receiving end sends an ACK message to notify the sending end to stop sending this block; after receiving the ACK, the sending end stops sending this block and sends the next block of data, and repeats the above process. Among them, the coverage window overlaps the previous and next data to a certain extent. When encoding, the coverage part is also included in the encoding as part of this block.

Step 2: The method of the present invention adopts a coding strategy based on non-uniform sampling. That is, in coding, the original fountain code completely random sampling strategy after the certainty is improved, and a non-uniform sampling strategy based on the number of sampling times is proposed. This strategy can optimize the performance of the fountain code under the condition of short code length and reduce redundancy.

Specifically, the coding strategy based on non-uniform sampling of the present invention improves the decoding success rate by balancing the sampling times of each source code. At each sampling, the corresponding sampling probability is calculated by the number of times the current source code has been sampled, and then a decision is made whether to sample the source code. In the specific algorithm design, the higher the sampling times, the lower the calculated sampling probability, so that the newly generated code samples more source codes that have not been fully encoded. The process of the coding strategy based on non-uniform sampling is shown in Figure 2. In the figure, the role of the threshold d ^* is to remove the non-uniform sampling of low-degree codes, and only focus on the sampling optimization of codes with larger degrees, which can reduce the number of subsequent probability-based selections and speed up the coding rate. The formula for calculating the probability of being sampled is:

π(d _s )＝exp(-c*d _s )

Where π(d _s ) is the sampling probability, c is a constant and c＞0, which ensures 0＜π(d _s )＜1. d _s is the number of times each source code is sampled, and d _s needs to be continuously updated in the source code set after sampling.

Furthermore, in the improved sampling strategy, the main difference is to judge the degree value and the threshold d ^* , that is, first determine the degree value d through completely random sampling based on the degree function. When d≤d ^* , directly randomly sample d source codes from the source code set to complete the sampling; when the degree value d＞d ^* , select d source codes from the source code set in turn, and calculate the probability of adding to the encoding set according to the number of times it is sampled, and then decide whether to add it according to this probability, and repeat it continuously until d source codes are selected from the source code set. Then, according to the source code set, the code is generated according to the XOR encoding method in the fountain code. The specific content is: when the degree value d＞d ^* , set the counter λ＝0, and then use π( _ds )＝exp(-c* _ds ) to decide whether to use it for encoding, that is: generate a decision probability p～U(0,1), and then make a judgment. If p＞π( _ds ), reselect the source code and calculate the sampling probability; if p≤π( _ds ), retain the source code and set the sampling number λ＝λ+1, and then execute the judgment of the sampling number λ; if λ＜ _ds , reselect the source code to calculate the sampling probability. If _λ≥ds , end the sampling.

Since π(d+1)＜π(d), the source code with fewer current encoding times has a higher probability of being sampled, and is sampled and encoded first. After all encoding is completed, the smaller the difference in the number of sampling times of each source code, the more conducive it is to decoding. Compared with the original encoding, after improving the sampling strategy, the average number of times the source code is encoded is significantly increased, and the difference between them is smaller, and the encoding is fairer. From the decoding end, the average impact of the lost part of the encoding is small.

Step 3: Based on the improved fountain code, the source data is continuously sent. According to the division method in step 1, a covering window transmission method is adopted. Each segment of data is dependent on each other. The decoded data of the previous block can be added to the decoding as the pre-information of the next block of data to improve the coding efficiency, and based on the forward error correction capability of the fountain code, reliable transmission is guaranteed without the ARQ mechanism. The code elements encoded by the fountain code are sent in blocks. The receiving end only retains the block information and coding information of the current data packet. It does not request retransmission for channel packet loss caused by transmission, but continuously attempts to decode based on the received packets. When the receiving end completes the decoding, it stops receiving the data packets of this block. The sending end sends the next data block to achieve continuous data transmission. That is, according to the division method in step 1, a covering window transmission method is adopted, so that each data segment is dependent on the head and tail, and the data of the previous decoded block is added to the decoding as the pre-information of the next block of data, and the code elements after block encoding are sent using the fountain code; the receiving end only retains the block information and encoding information of the current data packet, and continuously attempts to decode according to the received packets. When the receiving end completely completes the decoding, the receiving end stops receiving the data packet of this data block, and the sending end sends the next data block to achieve continuous data transmission. Preferably, in the covering window transmission method, the source code of the previous successfully decoded block is added to the next block of encoding as the pre-information of the decoding of this block, that is, it is added to the decoding as a node with a degree of 1 during decoding.

In the actual implementation of the method of the present invention, the design of the transmission message starts from the basic UDP protocol message format, and combines the information required for LT code encoding and decoding. The present invention redesigns the transmission message, which mainly adds the identification of the encoding block and the identification of the source code packet. As shown in Figure 3, the figure describes the message format of the reliable transmission protocol. In the specific format design, the first part is the original header information in the UDP protocol, including: source port (Source Port), destination port (Destination Port), message length (Length) and checksum (Checksum). The length field indicates the number of bytes in the UDP message segment, including the header and data, and the header information to ensure RDT is also included in the protocol design. The checksum is used by the receiver to check and check whether an error has occurred in the message segment. For the protocol that uses fountain codes to implement RDT, the present invention adds two fields, namely: source block index (Fountain Block Index) and encoding packet index (Encoding Source Index). The source block index identifies the position of the data packet generated by the encoding in all data blocks, which is convenient for the receiving end to recover all data based on the index, occupying 16 bits, and the maximum value is 65535. The code packet index identifies the source code used to encode this packet in the source code block, and uses 0/1 to indicate whether the source code packet is used. In order to ensure field alignment, a reserved field (Reserved Field) is added, and it can be used as an extension of the source code block index when transmitting longer data. The rest of the message is the application layer data (Message), and the specific size is determined by the packaging size of the upper layer application.

The data length occupied by the coding packet index needs to be adjusted according to the length of the actual coding block. As shown in Figure 4, it shows the case when 8 bytes are used to identify the source code packet. 8 bytes can identify the maximum length of the source code block of 64. In the actual design, it is necessary to estimate the length of the source code block according to the delay requirements and set the length of the source code packet index field. In the figure, a bit of 0 indicates that the source code with the serial number is not used during encoding, and a bit of 1 indicates that the source code is used during encoding. The figure shows that the source code with subscripts 0, 8, 24, and 47 in the code block is used for encoding, and the degree of encoding is 4.

From the perspective of protocol stack design, the reliable transmission method in the present invention uses UDP as the transport layer protocol and uses fountain codes to implement forward error correction. The combination of the two replaces the reliable transmission mechanism in the original network, as shown in Figure 5. Other fixed parts of the protocol stack such as the MAC layer and physical layer are omitted in the figure. At the sending end, the data sent from the transport layer is added with specific header information according to the encoding information, and then forwarded to the IP layer for transmission. At the receiving end, the data parsed by the IP layer is first parsed according to the fountain code encoding information in the packet, and corresponding decoding is performed until the current encoding block is successfully decoded and transmitted to the upper layer protocol. The transport layer protocol in the network uses the UDP protocol. The network IP layer protocol does not require a special specified protocol, and different protocols can be used according to the specific network architecture.

During the network data transmission process, each layer of protocol will add different packet headers according to the requirements of the protocol of this layer. Figure 6 describes the data packaging process involved in the protocol of the present invention. After the application layer data is generated, it is sent to the corresponding port. In the reliable transmission protocol: the UDP protocol of the transport layer adds the UDP message header to the data, and adds the relevant message headers for implementing fountain code encoding and decoding, and then transmits it to the lower layer routing protocol; after the IP layer receives the data, it also adds the header information and sends it. At the receiving end, the data is parsed layer by layer through the opposite process.

As shown in Figure 7, the upper layer service generates specific transmission data and stores it in the transmission buffer. Since the fountain code encoding requires all the information in the known packet, the data cannot be encoded and sent immediately after it is generated. When the data in the buffer meets the transmission requirements, the corresponding packet is read into the encoding buffer. The encoder regards the data in the buffer as a packet and starts to continuously generate and transmit the code. The receiving end's buffer is also divided into a receiving buffer and a decoding buffer. The receiving end first saves the successfully received data packet in the receiving buffer. The decoding buffer reads the corresponding data packet from the buffer to the decoding buffer according to the block sequence number that needs to be decoded currently, and starts decoding. Since the decoding result of the fountain code is out of order, the decoding end constantly checks the status of the current successfully decoded data. When a complete data block is detected to be successfully decoded, the controller arranges all data packets according to the index and uploads them to the upper layer receiving end. After the receiving end verifies that the information is correct, it sends a signal and clears the decoder buffer. When using the coverage window transmission, it is necessary to keep the data packet corresponding to the end of the previous packet in the buffer according to the set coverage ratio, and finally send ACK information to the sending end. The sending end ends the current group transmission according to the ACK information, and the controller clears the encoding buffer and attempts to send the next group according to the status of the sending buffer.

The method of the present invention improves the performance of the decoding end through a continuous transmission strategy based on a covering window; by dynamically updating the sampling probability according to the sampling number during the encoding process, the uniformity of the number of source code samplings is improved, and the redundancy of the short code and long fountain code is reduced.

The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed by the present invention, and these modifications or substitutions should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims (5)

1. A continuous reliable transmission method based on a coverage window and non-uniformly sampled fountain codes, comprising:

1) Dividing the data blocks of the transmitted data according to the time delay requirement;

2) Equalizing sampling times of each source code based on a coding strategy of non-uniform sampling, and generating codes according to a coding mode of fountain codes;

3) Based on the code transmission source data generated in the step 2), adopting a transmission mode of a coverage window to carry out continuous transmission according to the division mode of the step 1);

the specific content of the step 2) is as follows:

Firstly, sampling a degree value d from a degree function, comparing the degree value d with a threshold d ^*, and directly sampling from a source code set when the degree value d is less than or equal to d ^*, otherwise, sequentially selecting d source codes from the source code set, calculating the probability of adding the code set according to the sampled times, then determining whether to add according to the probability, repeating continuously until d source codes are selected from the source code set, and generating codes according to the XOR coding mode in fountain codes according to the source code set;

The formula for calculating the sampled probability is:

π(d_s)＝exp(-c*d_s)

Wherein pi (d _s) is the sampling probability, c is a constant and c is more than 0, which ensures that 0 < pi (d _s) < 1, which is the sampling frequency of each source code;

When the degree value d is larger than d ^*, the counter lambda=0 is made to generate decision probabilities p-U (0, 1), then judgment is carried out, if p is larger than pi (d _s), the source codes are selected again, and the sampling probability is calculated; if p is less than or equal to pi (d _s), reserving source codes, enabling sampling times lambda=lambda+1, and then executing judgment of the sampling times lambda; if lambda < d _s, the source code is selected again to calculate the sampling probability, and if lambda is more than or equal to d _s, the sampling is finished.

2. The continuous reliable transmission method based on the coverage window and the non-uniform sampling fountain codes as claimed in claim 1, wherein the specific contents of the step 1) are as follows:

Firstly, dividing the transmission data into data blocks with the same size, and then independently encoding each divided data block and then carrying out block transmission; after the transmission of each data block is finished, the receiving end sends an ACK message to inform the sending end to stop sending the data block; after receiving the ACK, the sending end stops sending the data block, and sends the next data block, and repeats the process; wherein, the cover window overlaps the front and back data, and when coding, the cover part is added into the coding as a part of the data block.

3. The continuous and reliable transmission method based on the coverage window and the non-uniform sampling fountain codes according to claim 1, wherein in the sampling times of each source code are balanced based on the coding strategy of non-uniform sampling, the sampling times of the current source code are utilized to calculate the probability of whether to add the sampling, whether to add the coding is determined according to the calculated probability, and then the process is repeated until the number of the source codes meeting the requirement is sampled.

4. The continuous reliable transmission method based on the coverage window and the non-uniform sampling fountain codes as claimed in claim 1, wherein the specific contents of the step 3) are as follows:

Based on the code transmission source data generated in the step 2), adopting a transmission mode of a coverage window according to the division mode in the step 1), enabling each section of data to depend from head to tail, adding the data of which the decoding is completed in the last block as the pre-information of the next block of data to be decoded, and transmitting code elements after the fountain code block coding; the receiving end only keeps the blocking information and the coding information of the current data packet, and continuously tries to decode according to the received packet, when the receiving end finishes decoding completely, the receiving end stops receiving the data packet of the data block, and the transmitting end transmits the next data block, so that continuous data transmission is realized.

5. The continuous and reliable transmission method based on the coverage window and the non-uniform sampling fountain codes as claimed in claim 4, wherein in the transmission mode of the coverage window, the source code of the successful decoding of the previous block is added to the next block code, and the source code is used as the pre-information of the decoding of the current block, namely, the source code is added to the decoding as a node with the degree of 1 during the decoding.