CN103905152B - Using the effective throughput randomized optimization process of layer-span combined coding in erasure channel - Google Patents

Abstract

The invention discloses a random optimization method for effective throughput using cross-layer joint coding in the deletion channel. The final data is transmitted to the physical layer and then R-S code encoding is performed, and the jointly encoded data is sent to the deletion channel; (2) After the transmission of the deletion channel, the data arriving at the receiving end is first decoded by R-S code at the physical layer After the decoding is completed, the decoded data is sent to the application layer for Raptor code decoding. After the decoding is completed, the receiving end will feed back the decoding situation to the sending end, and the sending end will take the next action according to the information fed back by the receiving end; (3) Use the discrete random approximation algorithm to select the appropriate cross-layer joint code rate to optimize the deletion channel communication Effective throughput in the system; the present invention can realize the effective throughput in maximizing the erasure channel stably and efficiently.

Description

Stochastic Optimization Method for Effective Throughput Using Cross-layer Joint Coding in Erasure Channels

technical field

The present invention relates to the coding technology field of erasure channel (Erasure channel) communication system, in particular to an erasure channel communication system using cross-layer joint coding (using fountain code as the forward error correction scheme of the application layer, and using Reed-Solomon The effective throughput stochastic optimization method is implemented by channel coding.

Background technique

The erasure channel is a general communication channel model used in coding theory and information theory. The information sent through the erasure channel is either received by the receiving end without error, or it is not received by the receiving end (called erasure). The deletion channel was proposed by Peter Elias in 1955, but it has been regarded as a theoretical channel model until the emergence of the Internet as a model in the real world of the deletion channel and its rapid development 40 years later made people realize that the deletion channel importance.

A common way to communicate in an erased channel is to use a feedback channel from the receiver to the sender to control the retransmission of erased channel packets. For example, the receiver may send back information identifying lost packets that will be retransmitted later by the sender. The receiver may alternately send an acknowledgment for each packet that has been correctly received. The sender keeps track of which packets have been acknowledged and which packets need to be retransmitted until all packets are acknowledged. These simple retransmission protocols have an advantage: they do not need to consider the channel erasure probability f. However, for many situations, these protocols are too wasteful of channel resources, especially in the broadcast erasure channel, and according to Shannon's theorem, there is no need for a feedback channel: whether we have feedback or not, the capacity of the forward channel is already reached (1-f) L bits (L is the total number of bits of information sent).

In order to solve the problem of resource waste, a kind of coding-based transmission solution is proposed. The original data is sent after being encoded with a linear error correction code. If some data is lost during transmission, it is possible to use erasure error correction algorithm to recover the lost data. Elias gives the capacity of a binary erasure channel with erasure probability p equal to 1-p. He further proved that random codes with a code rate arbitrarily close to 1-p can be successfully decoded on this channel with an exponentially small error probability by using the maximum likelihood decoding (ML) algorithm. But in the case of deleted channels, ML decoding of linear codes is equivalent to solving a system of linear equations. The solution of a system of linear equations can be done in polynomial time using Gaussian elimination, but when the code length is long, Gaussian elimination is not fast enough. RS codes can partially compensate for the inefficiency of random codes, but the decoding time is quadratic in the number of dimensions, which is too long for many applications. For Turnado codes with a linear time encoding/decoding algorithm, the encoding/decoding algorithm's running time is proportional to the block length, and if the channel rate is slow, the encoding/decoding algorithm will also be slow, and they are highly irregular for the base graph distribution of weights. LDPC code (Low Density Check Code) has good performance close to the Shannon limit, and the decoding complexity is low, but its encoding complexity is too high.

For linear error-correcting codes, the most important feature is to be able to correct as many errors as possible, while ensuring the complexity and ease of encoding and decoding algorithms. However, none of the traditional block coding methods mentioned above can satisfy these two important characteristics of error correction codes at the same time. Moreover, when data is sent from one sender to multiple receivers (the erasure probability of the erasure channel between the sender and each receiver is different), the defects of traditional block coding are more obvious. In a typical application, their code rate is detected by the sending end or the receiving end, so that the sending end can have a reasonable estimate of the current erasure probability of the channel and adjust the code rate accordingly. But if the number of receivers is large, or in the case of satellite/wireless transmissions, the receiving characteristics of the receivers change suddenly and significantly, it is impractical to track the loss rate of each receiver. The sender is then forced to assume the worst loss rate for all receivers. If the actual loss rate is small, it will bring unnecessary burden to the network. On the contrary, if the actual loss rate is greater than the pre-allocated, it will also lead to unreliable communication.

Fountain code, because of its code-free characteristics, does not need to know the probability of erasure of the channel, as long as enough data is sent to ensure that the receiving end receives a certain number of encoded packets, the original information can be reconstructed, and it has linear time encoding and decoding Complexity and other advantages, it can solve the problems that the above-mentioned traditional block coding methods cannot solve, so it becomes the focus of attention in the erasure channel. The present invention uses the Raptor code in the fountain code (the most effective type of code in the fountain code) as a forward error correction scheme of the application layer in the cross-layer coding, so as to provide end-to-end reliable transmission of the application layer.

Since the effect of using forward error correction coding alone in the application layer or physical channel is not ideal, the present invention considers the use of a cross-layer coding scheme to perform application layer coding and physical channel coding at the same time, combining the advantages of the two to greatly improve communication. the quality of. The channel coding scheme in the cross-layer coding is Reed-Solomon (R-S for short) code. The Reed-Solomon code reaches the Singleton boundary and is the most widely used code in practice, and the length of its code block is fixed. The reason why the fixed code rate coding is selected as the channel coding scheme is that the channel coding/decoding is processed by the chip, and the coding method with a fixed code rate is more efficient.

The effective throughput of the erasure channel is an important index to measure the channel utilization of the erasure channel. Effective throughput refers to the effective amount of information transmitted per unit time. After adopting the coding scheme of cross-layer coding, the effective throughput is a function of the verification rate of the Raptor code and the code rate of the Reed-Solomon code as independent variables, and its value will change with the channel state of the deleted channel. Generally speaking, before communication, we cannot predict the status information of the deleted channel, and it may change randomly. Under random channel conditions, in order to optimize the effective throughput, the most direct method is to make statistics on the state information of the channel first, and then take the average value after a long enough time of statistics. Choose the best check rate and code rate in the possible value space of Reed-Solomon code rate to maximize the effective throughput. One of the biggest drawbacks of this method is that the statistical time for channel state information needs to be long enough (theoretically infinite) to be accurate, so it takes too much time and is not practical. In order to optimize the effective throughput as the channel state information varies randomly, the present invention uses a sampling discrete random approximation algorithm to solve the problems existing in the direct method. This method will iteratively solve according to randomly selected channel state information, and the effective throughput will eventually converge to the optimal value after sufficient iterations.

Contents of the invention

The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art, to propose a forward error correction scheme using Raptor codes as the application layer to improve the robustness and reliability of transmission, and to use Reed-Solomon as channel coding at the same time A cross-layer joint coding scheme to reduce the probability of erasure in erasure channels. The joint coding scheme can restore the original sent data packets with a high probability through the coding of the application layer, providing reliable end-to-end transmission of the application layer, and reducing the error probability of the physical channel through channel coding to reduce the need for the application layer to send The number of encoded packets and the transmission delay, and a discrete random approximation algorithm is used to optimize the effective throughput in the erasure channel.

The purpose of the present invention is achieved through the following technical solutions: the effective throughput random optimization method of cross-layer joint coding adopted in the deletion channel is characterized in that, comprising the following steps:

(1) In the deletion channel communication system, after the sender gets the original data block delivered by the application process, determine the appropriate packet size L _p , divide the original data block into N _in data packets, and then use the Raptor code to compare the N _in original data packets. The data packet is encoded; the encoded data packet is passed down through the transport layer, network layer, and data link layer, and each layer that passes through adds a corresponding header and tail to each data packet, and finally reaches the physical layer; The layer performs RS code encoding on all data packets transmitted from the data link layer by selecting an appropriate code rate R, and sends them to the deletion channel after encoding is completed.

(2) After the transmission of the deleted channel, the data arriving at the receiving end is first decoded by R-S code at the physical layer of the receiving end, and after the decoding is completed, the decoded data is transmitted to the application layer for Raptor code decoding. After the decoding is completed, the receiving end will feed back the decoding situation to the sending end, and the sending end will take the next step according to the information fed back by the receiving end;

(3) Use the discrete random approximation algorithm to select the appropriate cross-layer joint code rate to optimize the effective throughput in the erasure channel communication system. The algorithm steps are as follows:

① According to the code rate R of the R-S code and the channel state information, calculate the deletion probability f of the deleted channel; then analyze the forward error correction using the Raptor code in the application layer, and obtain the probability p(ρ) of the Raptor code decoding failure; finally , in combination with the definition of effective throughput, the effective throughput is the effective amount of information transmitted per unit time, and the calculation formula y=y(ρ, R) of effective throughput is obtained, with the verification rate ρ of Raptor code and the R-S code The code rate R of is an independent variable, and changes with the change of the channel state;

② Define the verification rate of the Raptor code and the code rate of the R-S code as the cross-layer joint code rate, denoted as θ=(ρ, R), there are M possible discrete values of the verification rate ρ of the Raptor code, and the R-S code There are N possible discrete values of the code rate R, and the two together constitute N×M possible discrete values, so the size of the feasible search field Θ for all joint code rates is |Θ|=N×M;

③ After obtaining the calculation formula y=y(ρ,R) of effective throughput, use discrete random approximation algorithm to solve iteratively, and use i to represent the number of iterations; for iteration i=0, randomly select a joint code from the feasible search field rate as the initial joint code rate θ[0], and set θ[0] as the initial optimal joint code rate Corresponding to the jth element of the feasible search domain Θ; the state probability vector of all joint code rates is initialized as π[0]=e _θ[0] , the state probability vector is the same size as the feasible search domain Θ, and each component corresponds to The frequency at which each joint code rate in Θ is iteratively accessed by the algorithm, where e _θ[0] is a unit vector, the jth element is 1, and the rest of the elements are 0;

④ In the i-th iteration, the selected joint code rate θ[i], combined with the current channel state information, calculates the corresponding effective throughput y(θ[i]) according to the formula obtained in step ①;

⑤Uniformly and randomly select another joint code rate in the feasible search domain Θ Combined with the current channel state information, the corresponding effective throughput is calculated by the formula obtained in step ①

⑥ comparison and y(θ[i]) determine the joint code rate of the next iteration, if Then set the joint code rate of the next iteration Otherwise, let θ[i+1]=θ[i];

⑦ Let the iterative step size μ[i]=1/i, update the state probability vector of all joint code rates π[i+1]=π[i]+μ[i+1](e _θ[i+1] -π[i]), where e _θ[i+1] is a unit vector, and the position of 1 corresponds to the index of θ[i+1] in the feasible search domain Θ;

⑧ Compare the component corresponding to the joint code rate θ[i+1] in the updated state probability vector with the original optimal joint code rate The size of the corresponding component determines the optimal joint code rate for the next iteration if Update the current optimal joint code rate on the contrary,

⑨ Complete one iteration, return to step ④ to continue the next iteration; after the set number of iterations, the optimal joint code rate is obtained.

Preferably, in the step (1), all data packets arriving at the physical layer will be re-divided into data packets of size K=N·R according to the code rate R and code length N selected by the R-S code, and each data packet is divided into R-S After the code is encoded into a data packet with a length of N, it is sent to the deletion channel one after another;

Preferably, the discrete random approximation algorithm proposed in the step (3) optimizes the effective throughput of the erasure channel communication system in the algorithm step ②, the cross-layer joint code rate θ is defined as the verification rate ρ of the fountain code and the code of the R-S code The combination of rate R, that is, θ=[ρ, R].

Furthermore, in step ⑥ of the algorithm for optimizing the effective throughput of the erasure channel communication system by the discrete random approximation algorithm proposed in step (3), by combining the effective throughput corresponding to the joint code rate of this iteration with the value in the feasible search domain The effective throughput corresponding to another joint code rate randomly selected in the above is compared, and the joint code rate with greater effective throughput is selected as the joint code rate for the next iteration.

Furthermore, the discrete random approximation algorithm proposed in step (3) optimizes the effective throughput of the erasure channel communication system in steps ⑦ and ⑧, update the state probabilities of all joint code rates according to the joint code rate selected in step ⑥ vector, and the joint code rate with the largest access frequency in the state probability vector is taken as the current optimal joint code rate. After sufficient iterations, the optimal joint code rate obtained in the last iteration will be the joint code rate that has been accessed the most times in the entire iterative process, which is the final optimal joint code rate.

Furthermore, in step 9 of the algorithm for optimizing the effective throughput of the erasure channel communication system by the discrete random approximation algorithm proposed in step (3), in order to ensure that the algorithm performs sufficient iterations to converge to the optimal value, the number of iterations is feasible 2 to 4 times the size of the search field is more appropriate.

Preferably, the feedback method can be further combined to automatically request additional data packet transmission, the method is as follows:

(S1) If the application layer at the receiving end can recover all the original data, that is, the decoding is successful, then reply an ACK signal to the sending end to confirm the successful reception.

(S2) If the receiving end fails to decode, then reply a NACK signal to the sending end to notify the sending end to generate and send more fountain code output data packets.

(S3) If the sending end has not received any feedback signal after the waiting period T, automatically generate and send more fountain code output data packets. If the sender waits for the timeout three times in a row, it means that the link has been disconnected and stops sending data.

Compared with the prior art, the present invention has the following advantages and effects:

(1) In the deleted channel communication system of the present invention, the receiving end adopts a cumulative confirmation method for confirming the successful data reception of the sending end, and only sends back an ACK signal to the sending end after the entire data block is successfully received, saving Channel resources, improve channel utilization.

(2) The present invention combines an effective index for measuring the communication quality of the deleted channel in practical applications, that is, the effective information amount (effective throughput) successfully transmitted per unit time is taken as the optimization target.

(3) The present invention adopts a joint cross-layer coding scheme using Raptor code as the forward error correction scheme of the application layer and R-S code as channel coding at the same time to improve the robustness and reliability of transmission in the erasure channel and reduce transmission delay , to maximize the effective throughput.

(4) In the method of the present invention, the discrete random approximation algorithm is used to effectively search for the optimal joint code rate of the effective throughput that changes randomly with the channel state information, so that the effective throughput of the random change can be obtained under this optimal joint code rate. maximize.

(5) In the present invention, using the discrete random approximation algorithm to search for the optimal joint code rate is in multiple cycles, and judging whether the joint code rate is the current The most accessed joint code rate in the loop. Because in a certain iteration process, the effective throughput calculated according to a randomly selected joint code rate is greater than the effective throughput corresponding to the joint code rate selected in the previous iteration, this joint code rate is used as the next The joint code rate to be accessed, otherwise, the joint code rate will not be accessed. Therefore, after the iteration is over, the joint code rate corresponding to the largest component in the state probability vector of all joint code rates is the joint code rate that has been visited the most times, that is, the final optimal joint code rate. This method is accurate and efficient. .

Description of drawings

Fig. 1 is a schematic diagram of the system of the present invention.

Fig. 2 is a flow chart of the present invention.

detailed description

The present invention will be further described in detail below in conjunction with the embodiments and the accompanying drawings, but the embodiments of the present invention are not limited thereto.

Example

This embodiment deletes the effective throughput random optimization method of cross-layer joint coding in the channel, including the following steps, wherein the possible discrete value space of the Raptor code verification rate in this embodiment is [0, 0.1, 0.2, .. .2], there are 21 values in total, and the possible value space of the RS code rate is There are 5, and all possible values of the Raptor code check rate and the RS code rate constitute 101 joint code rates to choose from. Therefore, the size of the feasible search domain is 101. In order to achieve sufficient iterations, the number of selection cycles this time is set to 300;

(1) As shown in Figure 1, after the sender in the deletion channel communication system obtains the original data block delivered by the application process, determine the appropriate packet size L _p , divide the original data block into N _in data packets, and then use the Raptor code Encode the N _in original data packets; the encoded data packets are passed down through the transport layer, network layer, and data link layer, and each layer that passes through adds a corresponding header and tail to each data packet, Finally, it reaches the physical layer; the physical layer performs RS code encoding on all data packets transmitted from the data link layer by selecting an appropriate code rate R, and sends them to the deletion channel after encoding is completed;

(2) After the transmission of the deleted channel, the data arriving at the receiving end is first decoded by R-S code at the physical layer of the receiving end, and after the decoding is completed, the decoded data is transmitted to the application layer for Raptor code decoding. If the application layer at the receiving end can recover all the original data, it will reply an ACK signal to the sending end to confirm successful reception. Otherwise, reply a NACK signal to the sender, informing the sender to generate and send more fountain code output data packets. If the sending end has not received any feedback signal after the waiting period T, more fountain code output data packets are automatically generated and sent. If the sender waits for the timeout three times in a row, it means that the link has been disconnected and stops sending data.

(3) Use the discrete random approximation algorithm to select the appropriate cross-layer joint code rate to optimize the effective throughput in the erasure channel communication system. The algorithm flow is shown in Figure 2, and the specific steps are as follows:

① At the application layer, use Raptor code to encode N _in source data packets with a check rate ρ, and obtain the corresponding number of encoded data packets: N _o =(1+ρ)N _in . In the physical layer, the RS code with the code rate R is used for channel coding. Given a code length of N, the number of original information bits used for channel coding in each coded packet is K=N·R. If the bit error rate of the deleted channel is obtained according to the channel state information p _b , the deletion probability can be calculated as The number of encoded data packets received by the application layer at the receiving end is N _r =(1-f)N _o =(1-f)(1+ρ)N _in . The application layer at the receiving end starts decoding after receiving the encoded data packet, and the probability of decoding failure can be expressed as:

The probability of decoding failure is determined by the number of encoded packets received by the receiver. The amount of useful information obtained by the receiving end after decoding is N _use (R,ρ)=(1-p _a )N _in L _p , where L _p is the length of a data packet at the application layer. The total transmission time T(ρ,R)=N _o L _f /RR _s , where L _f is the length of the data link layer frame, and R _s is the data transmission rate of the physical layer. According to the definition of effective throughput, we get the calculation formula of effective throughput

② Randomly select one of the feasible search domains of the joint code rate as the initial joint code rate And set θ[0] as the initial optimal joint code rate Suppose θ[0] corresponds to the 64th element of the feasible search domain Θ. Set the 64th element in the initial state probability vector π[0] of all joint code rates as 1, and set the other elements as 0.

③Calculation by combining the deletion probability f(0) obtained from the channel state information and the effective throughput calculation formula we obtained in ① When the joint code rate is θ[0], the effective throughput is f(θ[0]) ;

④ Randomly select another joint code rate calculate Compare and the size of f(θ[0]);

like order

like order

⑤The state probability vector of the new joint code rate where e _θ[1] is a unit vector, the position of 1 corresponds to the index 23 of θ[1] in the feasible search domain Θ, if After updating, the values of the 23rd and 64th components of π[1] are both 1/2, and the remaining components are all 0. Otherwise, the 64th component value of π[1] is 1, and the remaining values are still 0, which is the same as the value of the 0th iteration.

⑥ Compare the joint code rate θ[1] and the optimal joint code rate in π[1] The size of the corresponding probability, update the current optimal joint code rate selection;

like order

like order

⑦ Return to step ④, enter the second cycle, and randomly select a joint code rate θ calculated Compared with f(θ[1]), the state probability vector of the joint code rate and the optimal joint code rate selection are updated; the cycle is continued, and the state probability vector of the joint code rate and the optimal joint code rate selection are updated in each cycle. Ends after the set number of iterations has been run.

As a result of random optimization of effective throughput using the method of this embodiment, the optimal joint code rate after the last cycle is selected stably and efficiently, so that the effective throughput in the erasure channel reaches the maximum value.

The above-mentioned embodiment is a preferred embodiment of the present invention, but the embodiment of the present invention is not limited by the above-mentioned embodiment, and any other changes, modifications, substitutions, combinations, Simplifications should be equivalent replacement methods, and all are included in the protection scope of the present invention.

Claims (8)

1. using the effective throughput randomized optimization process of layer-span combined coding in erasure channel, it is characterised in that including following Step：

(1) after the original data block that the transmitting terminal process of being applied is paid in erasure channel communication system, it is determined that suitably Bag size L_p, original data block is divided into N_inIndividual packet, then uses Raptor yards to this N in application layer_inIndividual initial data Bag is encoded；Packet after coding is passed down through transport layer, Internet, data link layer, and every layer of process is all Each packet adds corresponding head and afterbody, finally reaches physical layer；Physical layer is by selecting suitable code check R to data The entire packet that link layer is handed down carries out R-S code coding, is sent in erasure channel after the completion of coding；

(2) by the transmission of erasure channel, the data for reaching receiving terminal first carry out R-S code decoding in the physical layer of receiving terminal, solve Decoded data are passed into application layer after the completion of code carries out Raptor yards of decoding；After completing decoding, the feelings that receiving terminal will be decoded Condition feeds back to transmitting terminal, and transmitting terminal will take the action of next step according to receiving terminal feedack；

(3) using effective in the suitable layer-span combined optimization of rate erasure channel communication system of Discrete Stochastic approximate algorithm selection Handling capacity, comprises the following steps that：

1. the code check R and channel condition information according to R-S code obtain the probability of erasure f of erasure channel；Then application layer is adopted Forward error correction is carried out with Raptor yards to perform an analysis, obtain the Raptor yards of Probability p of decoding failure (ρ)；Finally, with reference to effectively handling up The definition of amount, the effective throughput is the effective information of the transmission of unit time, obtains the computing formula y of effective throughput =y (ρ, R), the code check R with Raptor yards of check rate ρ and R-S code changes as independent variable with the change of channel status；

2. the check rate of Raptor yards of definition and the code check of R-S code are designated as θ=(ρ, R), Raptor yards as layer-span combined code check The discrete value that is possible to of check rate ρ have M, the discrete value that is possible to of the code check R of R-S code has N number of, and both are common N × M possible discrete value is constituted, therefore the size of the feasible region of search Θ of all unicode rates is | Θ |=N × M；

3. after obtaining the computing formula y=y (ρ, R) of effective throughput, solution is iterated with Discrete Stochastic approximate algorithm, uses i Represent iterations；The i-th=0 iteration, randomly chooses a unicode rate as initial joint code from feasible region of search Rate θ [0], and θ [0] is set to initial optimal joint code checkJ-th yuan corresponding to feasible region of search Θ of θ [0] Element；The state probability vector for initializing all unicode rates is π [0]=e_θ[0], the state probability vector and feasible region of search The same sizes of Θ, the frequency that each component is accessed corresponding to each unicode rate in Θ by algorithm iteration, wherein e_θ[0]It is one Unit vector, j-th element is 1, and remaining element is 0；

4. in ith iteration, the unicode rate of selection is θ [i], with reference to current channel condition information, is 1. obtained according to step Formula calculate corresponding effective throughput y (θ [i])；

5. uniformly random in feasible region of search Θ another unicode rate of selectionAnd with reference to current channel condition information, The formula 1. obtained by step calculates corresponding effective throughput

6. compareSize with y (θ [i]) determines the unicode rate of iteration next time, ifThen set The unicode rate for putting next iteration isOtherwise, θ [i+1]=θ [i] is made；

7. step size mu [the i]=1/i of iteration is made, state probability vector π [i+1]=π [i]+μ [i+1] of all unicode rates is updated (e_θ[i+1]- π [i]), wherein e_θ[i+1]It is a unit vector, 1 position corresponds to fingers of the θ [i+1] in feasible region of search Θ Mark；

8. the corresponding components of unicode rate θ [i+1] and former optimal joint code check in the state probability vector after updating are comparedIt is right The size of the component answered, it is determined that the optimal joint code check of next iterationIf Updating current optimal unicode rate isConversely,

9. an iteration is completed, step is returned to and is 4. continued next iteration；By after the iterations for setting, obtaining optimal connection Close code check.

2. using the effective throughput randomized optimization process of layer-span combined coding in erasure channel according to claim 1, Characterized in that, in the step (1), reaching code check R and code length N that all packets of physical layer will be selected according to R-S code Again it is divided into the packet that size is K=NR, after each packet R-S code is encoded into the packet that length is N, sends out successively It is sent in erasure channel.

3. using the effective throughput randomized optimization process of layer-span combined coding in erasure channel according to claim 1, Characterized in that, in the step (3), layer-span combined code check θ is defined as the group of the check rate ρ of fountain codes and the code check R of R-S code Close, i.e. θ=[ρ, R].

4. using the effective throughput randomized optimization process of layer-span combined coding in erasure channel according to claim 1, Characterized in that, the step 2. in, the joint code that the code check of Raptor yards of check rate and R-S code is encoded as cross-layer Rate, determines feasible region of search.

5. using the effective throughput randomized optimization process of layer-span combined coding in erasure channel according to claim 1, Characterized in that, the step 6. in by by the corresponding effective throughput of the unicode rate of current iteration with feasible region of search In the corresponding effective throughput of another unicode rate that randomly selects be compared, joint of the selection with bigger effective throughput Code check as next iteration unicode rate.

6. using the effective throughput randomized optimization process of layer-span combined coding in erasure channel according to claim 1, Characterized in that, step 7., 8. in, the unicode rate according to step 6. middle selection update the state probability of all unicode rates to Amount, and using the unicode rate with largest access frequency in state probability vector as current optimal unicode rate, Suo Youlian Each unicode rate is changed to current by algorithm during each component of the state probability vector of conjunction code check has corresponded to feasible region of search The frequency that generation accesses；By the optimal joint code check that after abundant iteration, last time iteration is obtained will be in whole iterative process The most unicode rate of accessed number of times, as final optimal joint code check.

7. using the effective throughput randomized optimization process of layer-span combined coding in erasure channel according to claim 1, Characterized in that, step 9. in, to ensure that algorithm carries out abundant iteration to converge to optimal value, the iterations is searched for feasible 2~4 times of rope domain size.

8. using the effective throughput randomized optimization process of layer-span combined coding in erasure channel according to claim 1, Characterized in that, further combined with the method for feedback, automatic request excessive data bag transmission, specifically：

(S1) if receiving terminal application layer can recover whole original data, i.e. successfully decoded, then an ack signal is replied To transmitting terminal, confirm to be properly received；

(S2) if receiving terminal decoding failure, replys a NACK signal to transmitting terminal, transmitting terminal is notified to produce and send more Many fountain codes output data packets；

(S3) if transmitting terminal does not receive any feedback signal yet after waiting period T, automatically generate and send more sprays Spring code output data packet；If transmitting terminal waits time-out continuous three times, illustrate that link has been disconnected, stop sending data.