CN105846958A - Distributed system Raptor code transmission method specific to deep space communication - Google Patents

Abstract

The present invention provides a distributed system Raptor code transmission method oriented to deep space communication, comprising the following steps: step S1, performing system Raptor encoding _on the original information of the source end with code _lengths K1 and K2 respectively and sending them to the same Relay; step S2, the relay R stores the encoded symbols from the source end in the buffer area E ₁ and the buffer area E ₂ respectively, and then uses the DSRC algorithm to process the data and then sends it to the destination end; step S3, the destination end D pairs the receiving end Decode the received code symbols. The present invention designs a distributed system Raptor code transmission method for the scene of multiple detectors passing through the orbiter to the ground, proposes a simplified joint decoding scheme, theoretically analyzes and deduces the performance parameters of the DSRC scheme and its improved scheme, and Compared with the distributed rateless erasure correction scheme in the prior art, a 99% decoding success rate is obtained when the redundancy reaches 5%.

Description

Raptor Code Transmission Method for Distributed System Oriented to Deep Space Communication

technical field

The invention relates to a code transmission method oriented to deep space communication, in particular to a distributed system Raptor code transmission method oriented to deep space communication.

Background technique

With the diversification of deep space exploration tasks and the continuous extension of the detection range, the amount of data and business types that need to be transmitted is increasing day by day, especially in the face of the needs of multimedia data services such as images, videos, and voices for deep space exploration in the future. The point-to-point support measures based on physical quantity gains are unsustainable, and the use of relay, collaboration, and even networked transmission is an inevitable trend in the development of deep space communications in the future; the "Space Interconnection Strategy Group" established by various aerospace organizations in the world including China (SISG), established the Delay/Disruption Tolerant Networks (DTN) protocol system as the core, around the relevant standards and recommendations of the Space Data System Consultative Committee (CCSDS), integrated measurement and control resources at all levels, and established a deep space Probing the communication network.

The currently running Mars exploration missions and related research have shifted from single-satellite and single-user to relay-satellite networking. The "Curiosity" Mars Exploration Rover that landed in October 2012 returned most of the video and image data to the earth through the "Odyssey" and "Mars Reconnaissance Orbiter", and the relay communication rate reached more than 40 times that of the direct link ; However, even with multi-hop network transmission, the single-hop link distance of interstellar communication in deep space is still many orders of magnitude larger than the distance of satellite communication in near-Earth, and the existing channel coding that has approached the Shannon limit cannot correct all bit errors. However, the 3-20 minute propagation delay between Mars and Earth makes the confirmation retransmission method of the existing transmission protocol very inefficient.

Based on this, CCSDS proposes to use Long Erasure Code (LEC) to provide forward erasure correction capability for data packets in the transport layer. The packet is erasure coded, so that the destination D can use the erasure group to restore the deleted information packet, so as to reduce the number of feedback retransmissions and improve transmission efficiency. At present, the research on LEC is in its infancy, and the digital fountain, as a code-free forward erasure grouping technology with low encoding and decoding complexity and can approach the Shannon limit with any probability, has been proved to be a suitable method for deep space communication. The LEC erasure coding scheme. In addition, the relay orbiter adopts an asynchronous store-and-forward mechanism, which can use the received data in the cache to design an asynchronous LEC code without considering the synchronization of the link layer, and provide distributed intermediate data for multiple measured star surface nodes. continue transmission.

Contents of the invention

The technical problem to be solved by the present invention is to provide a distributed system Raptor code transmission method that can improve the success rate of decoding under deep space communication.

In this regard, the present invention provides a distributed system Raptor code transmission method for deep space communication, comprising the following steps:

Step S1, system Raptor codes the original information at the source end with code lengths K ₁ and K ₂ respectively and sends them to the same relay;

Step S2, the relay R stores the encoded symbols from the source end in the buffer area _E1 and the buffer area _E2 respectively, and then uses the DSRC algorithm to process the data and then send it to the destination end;

In step S3, the destination D decodes the received encoded symbols.

A further improvement of the present invention is that, in the step S1, the system Raptor encoding process at the source end is to fill in zeros with K source symbols C, and generate L intermediate symbols C' through an intermediate generator matrix G constructed by a pseudo-random sequence , and satisfy the following formula: Among them, the intermediate generation matrix G includes S LDPC coded symbols G _LDPC , H Half coded symbols G _Half and K LT coded symbols G _LT , L=S+H+K, IS is a unit matrix of S rows and _S columns , 0 _{S × H} is the zero matrix of S rows and H columns, I _H is the unit matrix of H rows and H columns; S, L, H and K are all natural numbers, representing the rows and columns of the coding matrix; C' performs LT encoding to generate redundant coded packets, and the coded matrix is marked as R _LT ; then the original symbols and redundant coded symbols are sent in sequence. Among them, the formula Each parameter symbol in the square brackets represents a schematic structure of sub-matrix arrangement inside a matrix.

A further improvement of the present invention is that in the step S1, the information source W ₁ and the information source W ₂ at the information source end perform system Raptor encoding independently and then send their original symbols and redundant encoding symbols to the relay R, and store them in the _In buffer area E _{1 and} buffer area E _{2 ;} wherein, in step S2, the relay R _caches the original symbols, the redundant coded symbols of source W ₁ and source W ₂ are buffered respectively through the buffer area E _1,r and the buffer area E _2,r .

A further improvement of the present invention lies in that, in the step S2, the coded symbols corresponding to the source W ₁ and the source W ₂ reach the relay R via erasure channels with erasure probabilities P _wr1 and P _wr2 respectively.

A further improvement of the present invention is that, in the step S2, the relay R uses the DSRC algorithm to process data including the following sub-steps:

Step S201, forwarding the original symbols in the buffer area E _1,w or the buffer area E _2,w until all the buffered original symbols are forwarded;

Step S202, realizing distributed encoding and forwarding through the relay R.

A further improvement of the present invention is that in the step S201, if the cache of the buffer area E _1,w or the buffer area E _2,w is not empty, randomly select from the buffer area E _1,w or the buffer area E _2,w One original symbol is forwarded until all buffered original symbols are forwarded, then jump to step S202.

A further improvement of the present invention is that step S202 implements distributed encoding forwarding in any of the following three ways: Randomly select a redundant encoding symbol from the buffer area E _1,r or buffer area E _2,r to implement That is, store and forward; randomly select a redundant coding symbol from the buffer area E _1,r and the buffer area E _2,r to realize XOR forwarding; and randomly select from the buffer area E _1,w or the buffer area E _2,w An original symbol is selected, and a redundant coding symbol is randomly selected from the buffer area E _2,r or the buffer area E _1,r , and XOR forwarding is realized.

A further improvement of the present invention is that, in the step S3, the coded symbols are sent to the destination D through the RD section link with a deletion probability of P _rd , firstly, the joint matrix transformation process is used to simplify the processing, and then the remaining ununitable Stop set matrix for Gaussian decoding.

A further improvement of the present invention is that the step S3 includes the following sub-steps:

Step S301, initialize the state, and perform row elementary transformation and column elementary exchange on the decoding matrix;

Step S302, realizing matrix simplification through row transformation and transformation elimination;

In step S303, the unitization of the decoding matrix is realized, and the decoding ends.

A further improvement of the present invention is that, in the step S301, the state is initialized, and the row elementary transformation and the column elementary exchange are respectively performed _on the block A1 and the block A2 to simplify the matrix for the _first time, and then use the obtained _I1 and _I2 to R _{LT performs} the elimination process, in which, block A ₁ is the intermediate generation matrix G _{1 corresponding to the information source W 1 ,} block A ₂ is the intermediate generation matrix G 2 corresponding to the information source W ₂ , and I 1 is the intermediate generation matrix G ₂ corresponding to the information source W 2, and I ₁ is for the block A ₁ to perform The unit matrix obtained after elementary transformation simplification, I ₂ is the unit matrix obtained after block A ₂ is simplified after elementary transformation, R _LT is the redundant symbol obtained by random XORing the redundant symbols from two information sources In the step S302, U ₁ and U ₂ are divided into upper and lower parts, and U _1,low and U _2,low are changed into an upper triangular matrix through row elementary transformation, and U 1,low after transformation is utilized _{. low} and U _{2,low perform} row transformation and elimination on the blocks [0, B ₁ , 0, B ₂ ], and then select appropriate rows to add to U _1,low and U _2,low to obtain C ₁ and C ₂ , where U ₁ is a non-zero matrix with L rows and unknown number of columns generated after the first-stage elementary transformation of block A ₁ ; block U ₂ is the first-stage elementary transformation of block A ₂ After that, the generated non-zero matrix with L rows and unknown columns, U _1,low is the upper triangular matrix after the elementary transformation of U ₁ , U _2,low is the upper triangular matrix after the elementary transformation of U ₂ , divided into The block [0, B ₁ , 0, B ₂ ] is the non-zero matrix after the elementary transformation of RLT, and the sub-block C ₁ is the remaining sub-matrix after the elementary transformation of U ₁ to obtain the upper triangular matrix U _1,low . Block C ₂ is the remaining sub-matrix after performing elementary transformation on U ₂ to obtain the upper triangular matrix U _2,low ; in the step S303, the remaining M-2L rows are discarded, and the blocks are marked as A ₁₁ and A ₂₂ , and They are unitized respectively, and the decoding is completed, where M is the number of rows of the received joint decoding matrix, L is the code length of the systematic code determined by the input source code length K, and A ₁₁ and A ₂₂ are joint decoding Simplified matrix.

Compared with the prior art, the beneficial effects of the present invention are: based on the protocol framework of the delay-tolerant and fault-tolerant network (DTN), the orbiter is used to carry out multi-hop relay network transmission, and the design is designed for multiple detectors to pass through the orbiter to the ground. The distributed system Raptor code transmission method in the transmission scenario, proposed a simplified joint decoding scheme, theoretically analyzed and deduced the performance parameters of the DSRC scheme and its improved scheme, and simulated it with the existing distributed rateless erasure correction scheme In comparison, a decoding success rate of 99% is obtained when the redundancy reaches 5%.

Description of drawings

Fig. 1 is the Y-type deep space communication network schematic diagram of an embodiment of the present invention;

Fig. 2 is a schematic diagram of a decoding matrix of an embodiment of the present invention;

Fig. 3 is a schematic diagram of the joint decoding matrix transformation process in step S3 of an embodiment of the present invention;

Fig. 4 is a schematic diagram of performance analysis based on an AND-OR tree in an embodiment of the present invention;

Fig. 5 is the decoding progressive performance emulation diagram of the DSRC relay strategy of an embodiment of the present invention;

Fig. 6 is a decoding performance simulation diagram of changing the packet loss rate between the source and the relay R in the DSRC scheme of an embodiment of the present invention;

Fig. 7 is a decoding performance simulation diagram of changing the packet loss rate between the relay R and the destination D in the DSRC scheme of an embodiment of the present invention;

Fig. 8 is a performance comparison simulation diagram between the DSRC scheme of an embodiment of the present invention and the prior art;

FIG. 9 is a simulation diagram of decoding performance of a DSRC scheme under different parameters according to an embodiment of the present invention.

detailed description

Below in conjunction with accompanying drawing, preferred embodiment of the present invention is described in further detail:

This example provides a distributed system Raptor code transmission method for deep space communication, including the following steps:

Step S1, system Raptor codes the original information at the source end with code lengths K ₁ and K ₂ respectively and sends them to the same relay;

Step S2, the relay R stores the coded symbols from the source end in the buffer area _E1 and the buffer area E2 respectively, and then uses the DSRC algorithm to process the data and then send it to the destination end;

In step S3, the destination D decodes the received encoded symbols.

This example is aimed at the communication scenario where the surface detectors of the measured star jointly use the orbiter relay to return scientific data in the future. As shown in Figure 1, a Y with two sources, a single relay R and a single destination D is designed. type topology, where the source at the source end is also called a source node, the relay R is also called a relay node, and the destination D is also called a destination node; then combined with the DTN asynchronous store-and-forward mechanism, this example is aimed at deep space communication The network has designed a distributed system Raptor code transmission method, that is, the Distributed Systematic Raptor Coding scheme, referred to as the DSRC scheme; furthermore, in this example, considering the limited energy of the deep space probe and the storage capacity of the relay orbiter, a preset source code is proposed The number of symbols and the optimization scheme of relay network coding can effectively reduce the coding overhead; finally, the performance of the DSRC scheme described in this example is verified and analyzed through simulation.

Since the orbiter as a relay node has the characteristics of periodic interruption, the distributed erasure code scheme designed in the existing ground wireless communication is not suitable for the deep space communication network; in order to further improve the effective throughput in practical scenarios, this paper For example, the system Raptor code is proposed to be used at the source end to reduce the erasure correction processing overhead of the probe car and the relay; the Raptor code is an improved version developed on the basis of the LT code, and the LT code is a general-purpose digital fountain code.

In Figure 1, the working process of each node in the transmission process is as follows: the information source, information source W ₁ and information source W ₂ are independent of each other, and perform system Raptor encoding on the original information with code length K ₁ and K ₂ respectively And send; Relay R, code lengths K ₁ and K ₂ source W ₁ and source W ₂ coded symbols of the original information arrive at relay R through erasure channels with erasure probabilities P _wr1 and P _wr2 respectively, The relay R stores the encoded symbols of the source W ₁ and the source W ₂ in the buffer area E ₁ and the buffer area E ₂ respectively, and then uses the DSRC algorithm to perform operations including XOR, discarding, and forwarding, and then sends them to the destination; At the destination D, the coded symbols arrive at the destination D through the RD section link with the deletion probability P _rd . First, the joint matrix transformation process is used to simplify the process, and then Gaussian decoding is performed on the remaining stop set matrices that cannot be unitized.

In the step S1, the system Raptor encoding process at the source end is to fill in zeros with K source symbols C, then construct an intermediate generator matrix G through a pseudo-random sequence to generate L intermediate symbols C', and satisfy the following formula: Among them, the intermediate generation matrix G includes S LDPC coded symbols G _LDPC , H Half coded symbols G _Half and K LT coded symbols G _LT , L=S+H+K, IS is a unit matrix of S rows and _S columns , 0 _{S × H} is the zero matrix of S rows and H columns, I _H is the unit matrix of H rows and H columns; S, L, H and K are all natural numbers, representing the rows and columns of the coding matrix; C' performs LT encoding to generate redundant coded packets, and the coded matrix is marked as R _LT ; then the original symbols and redundant coded symbols are sent in sequence. The original symbols are also called system symbols. Among them, the formula Each parameter symbol in the square brackets represents the schematic structure of the sub-matrix arrangement inside a matrix, and belongs to the expression writing method of the sub-matrix in the matrix.

In this embodiment, different K corresponds to different L, such as K=16,17,18 corresponds to L is 24; when K=244,255,266, L is 264, this can be adjusted according to the actual situation; as shown in Figure 2, In the matrix of the first row, there are S rows and L columns in total; G _LDPC includes S rows and (LHS) columns; I _S is a unit matrix of S rows and S columns; 0 _S×H is a zero matrix of S rows and H columns . In the second row, there are H rows and L columns in total; G _Half includes H rows and (LH) columns; I _H is the unit matrix of H rows and H columns; in the third row, there are K rows and L columns in total; G _LT includes K rows, L columns; the configuration of the matrix G, such as the sub-matrix A ₁ or A ₂ of the decoding matrix A, as shown in FIG. 2 .

In step S1 described in this example, source W ₁ and source W ₂ at the source end perform system Raptor encoding independently and then send their original symbols and redundant coded symbols to relay R, and store them in buffer areas E ₁ and In the buffer area E ₂ ; wherein, in step S2, the relay R buffers the original symbols of the source W ₁ and the source W ₂ respectively through the buffer area E _{1, w} and the buffer area E _{2, w} , and through the buffer area E _1,r and the buffer area E _2,r buffer the redundant coding symbols of the information source W ₁ and the information source W ₂ respectively.

In step S2 in this example, the encoded symbols corresponding to source W ₁ and source W ₂ arrive at relay R via erasure channels with erasure probabilities P _wr1 and P _wr2 respectively.

In step S2 described in this example, the relay R uses the DSRC algorithm for data processing including the following sub-steps:

Step S201, forwarding the original symbols in the buffer area E _1,w or the buffer area E _2,w until all the buffered original symbols are forwarded;

Step S202, realizing distributed encoding and forwarding through the relay R.

Wherein, in the step S201, if the cache of the buffer area E _1,w or the buffer area E _2,w is not empty, then randomly select an original symbol from the buffer area E _1,w or the buffer area E _2,w for forwarding , until all cached original symbols are forwarded, then jump to step S202.

The step S202 implements distributed coding forwarding in any one of the following three ways: the first way is to randomly select a redundant coding symbol from the buffer area E _1,r or buffer area E _2,r to implement and store Forwarding, in the accompanying drawings 5 and 9 of the description, the first method is represented as method 1; the second method is realized by randomly selecting a redundant coding symbol from the buffer area E _{1, r} and the buffer area E _{2, r XOR} forwarding, in the accompanying drawings 5 and 9 of the specification, the second method is represented as method 2; and, the third method is to _randomly select an original Symbol, a redundant coding symbol randomly selected from the buffer area E _{2, r} or buffer area E _{1, r} , and then realize XOR forwarding. In the accompanying drawings 5 and 9 of the specification, the third method is expressed as the method 3.

It is worth mentioning that the DSRC scheme is approximately equivalent to expanding two decoding matrices with L columns into a joint decoding matrix with more than 2L columns.

After receiving M (M>2L) encoded symbols, the destination D first reconstructs the decoding matrix A, as shown in Figure 2, the joint decoding extends the decoding matrix, and the algorithmic complexity of Gaussian decoding increases to the original 8 times; therefore, this example also preferably proposes a simplified decoding algorithm for optimizing the block 0 matrix in step S3.

In step S3 in this example, the coded symbols are sent to the destination D through the RD link with the deletion probability P _rd , using a Gaussian decoding algorithm based on a stopping set. Specifically, the step S3 includes the following sub-steps:

Step S301, initialize the state, and perform row elementary transformation and column elementary exchange on the decoding matrix;

Step S302, realizing matrix simplification through row transformation and transformation elimination;

In step S303, the unitization of the decoding matrix is realized, and the decoding ends.

In said step S301, the initialization state, as shown in Fig. 3 (a), performs row elementary transformation and column elementary exchange respectively to sub-block A ₁ and sub-block A ₂ to simplify the matrix for the first time, obtain Fig. 3 (b), then use The obtained I ₁ and I ₂ perform elimination processing on R _LT to obtain Figure 3(c), in which, the block A ₁ is the intermediate generator matrix G ₁ corresponding to the information source W ₁ , and the block A ₂ is the information source W ₂ Corresponding intermediate generator matrix G ₂ , I ₁ is the identity matrix obtained after block A ₁ is simplified after elementary transformation, I ₂ is the identity matrix obtained after block A ₂ is simplified after elementary transformation, R _LT is the pair from two The generation matrix of redundant symbols obtained by randomly _XORing redundant symbols of sources _; in the step S302, U1 and U2 are divided into upper and lower parts, and U1 _{, low} and _U2 are divided into upper and lower parts by row elementary transformation _,low becomes an upper triangular matrix, and use the transformed U _1,low and U _2,low to perform row transformation and elimination on the block [0, B ₁ ,0, B ₂ ], and then select the appropriate row to add to U _1,low and U _2,low to obtain C ₁ and C ₂ , where U ₁ is a non-zero matrix with L rows and unknown number of columns generated after the first-stage elementary transformation of block A ₁ ; Block U ₂ is the non-zero matrix with L rows and unknown number of columns generated after the first-stage elementary transformation of block A ₂ , U _1,low is the upper triangular matrix after elementary transformation of U ₁ , U _{2 , low} is the upper triangular matrix after the elementary transformation of U ₂ , the block [0, B ₁ , 0, B ₂ ] is the non-zero matrix after the elementary transformation of RLT, and the block C ₁ is the elementary transformation of U ₁ Transformation obtains the upper triangular matrix U _{1, the remaining sub-matrix after low} , and the sub-block C ₂ is to carry out elementary transformation to U ₂ to obtain the upper triangular matrix U _{2, the remaining sub-matrix after low} ; in the step S303, discard the remaining M-2L rows, the blocks are marked as A ₁₁ and A ₂₂ , and are unitized respectively, and the decoding ends, where M is the number of rows of the received joint decoding matrix, and the M is determined by the source W ₁ and the source The encoding symbols generated by W ₂ , after the corresponding generator matrices G ₁ and G ₂ are filled with zeros, are generated in the first stage of the relay algorithm, so it is 2L, plus the XOR redundant symbols generated in the second stage, that is, R _LT , the number of rows>0; so M>2L; L is the code length of the systematic code determined by the input source code length K; A ₁₁ and A ₂₂ are simplified matrices after joint decoding, which is similar to that in Figure 3(c) The matrices in are indistinguishable. If the decoding matrix cannot be unitized, the decoding fails.

This example is based on the Delay Tolerant Network (DTN) protocol framework, using the orbiter for multi-hop relay network transmission, and designing a distributed system Raptor code transmission method for the scenario where multiple detectors are transmitted from the orbiter to the ground. A simplified scheme of joint decoding is proposed, the performance parameters of the DSRC scheme and its improved scheme are deduced through theoretical analysis, and compared with the existing distributed rateless erasure correction scheme through simulation, 99% is obtained when the redundancy reaches 5%. % decoding success rate.

Below, this example analyzes the complexity of the DSRC scheme: the average number of XOR operations required for encoding and decoding corresponding to each symbol of the system Raptor code is: In this formula, N represents the number of coding symbols, and K represents the code length.

In the DSRC scheme, for source W ₁ and source W ₂ whose code lengths are K ₁ and K ₂ respectively, source encoding generates N ₁ and N ₂ encoded symbols, and the total number of XOR operations in the encoding process is: J _enctotal ＝N ₁ ·J _enc (N ₁ ,K ₁ )+N ₂ ·J _enc (N ₂ ,K ₂ ), the network coding operation of the relay is only performed on redundant coding symbols, and the required number of XOR operations is : J _NC ＝min{(1-P _wr1 )·(N ₁ -K ₁ ),(1-P _wr2 )·(N ₂ -K ₂ )}, where P _wr1 and P _wr2 represent W ₁ -R The deletion probability of segment link and W ₂ -R segment link. Correspondingly, the destination has: J _dectotal =K ₁ ·J _dec1 +K ₂ ·J _dec2 =K ₁ ·J _dec (M ₁ ,K ₁ )+K ₂ ·J _dec (M ₂ ,K ₂ ), where, M ₁ and M ₂ represent the number of encoded symbols of source W ₁ and source W ₂ respectively received by destination D, when K ₁ =K ₂ =K, P _wr1 =P _wr2 =p, then N ₁ ≈N ₂ =N, M ₁ ≈M ₂ =M.

For distributed system Raptor codes, the decoding operation is only required when system symbols are lost. Compared with the total number of XOR operations J _enctotal of the codec and the total number of operations J _dectotal of the destination, the required number of XOR operations J _NC is almost negligible, and the approximate total number of XOR operations is: J _total = (1-( 1-p) ^2K )(J _enctotal +J _NC +J _dectotal )≈2·(1-(1-p) ^2K )(10·K·N+4.5·K ² ).

It can be seen from the above analysis that the DSRC scheme has an approximately linear O(K) encoding complexity and O(K ² ) decoding complexity.

Next, this example also proposes an improved distributed system Raptor code transmission method, that is, an improved DSRC scheme. First, the decoding failure probability of the DSRC scheme is analyzed. Further analysis of the DSRC scheme: the source only generates a fixed number of coded symbols, and the relay guarantees data recovery at the destination by network coding or forwarding the received coded symbols.

First of all, it is necessary to determine the number of source coded symbols. The W ₁ -R segment link and the W ₂ -R segment link can be regarded as two point-to-point communication links. The empirical formula between the number of symbols m and the code length K: Then if the relay decoding failure probability ε is known, use the formula The number of coded symbols m _th that need to be received when the success rate of relay decoding reaches (1-ε) can be estimated; if the packet loss rate P _wr of the WR segment link is predicted, the number of coded symbols sent by each source can be determined: m _th /(1-P _wr ). Theoretically, using these encoded symbols that contain all the original information to perform appropriate operations can ensure successful decoding at the destination.

In the case of knowing the channel erasure probability in advance, the number of encoded symbols sent by the source end is determined, then the probability of relay decoding failure at this time is: The lower bound of the decoding failure probability that can be achieved by relaying through the three selective network coding methods in step S202 is P _fr .

when hour, Have At the same time, the probability of decoding failure at the destination can be expressed in the following form: P _ft =1-(1-P _f (M ₁ ,K ₁ ,δ _i ))·(1-P _f (M ₂ ,K ₂ ,δ _i )). in: M ₁ and M ₂ represent the number of system symbols received from sources W ₁ and W ₂ ; M ₀ represents the number of symbols generated by the relay network coding after the RD section deletes the channel and the destination receives the number. Because of the uniform randomness of network coding, it is approximately considered that M ₁ =M ₂ ; and δ _i (i=1, 2, 3) respectively represent the influence of three selective relay network coding methods on decoding performance.

The following is an analysis of the performance limit of the DSRC scheme based on the AND-OR tree. The most commonly used tool for analyzing the decoding performance of the fountain code is the AND-OR tree analysis. As shown in Figure 4, an AND-OR tree with a depth of 2l+1 is defined. The depth of the root node is defined as 0, and the child nodes of each layer are expanded downwards in turn. Nodes at even layers (0,2,4,...,2l) correspond to input nodes, called "OR" nodes, and nodes at odd layers (1,3,5,...,2l-1) The corresponding output node is called an "AND" node. For an LT code whose output degree distribution is Ω(x), when the edge distribution of the input node obeys the binomial distribution (1/K,αK), and K→∞, the input degree distribution approximately obeys the Poisson distribution exp(α(x -1)), α is the average degree of its input nodes. The relationship between output side distribution ω(x) and Ω(x) is: ω(x)=Ω'(x)/Ω'(1). Define the decoding failure probability of a certain input node in the initial stage of decoding as y ₀ , and after l decoding iterations, the decoding failure probability y _l of this node is related to ω(x) and Ω(x) by and or Tree theorem formula:

Assuming that the decoding redundancy ratio of the destination D is ε, then the relationship between α and the average degree of the output node Ω'(1) and ε is: α=(1+ε)Ω'(1), the formula can be written as:

For source W ₁ and source W ₂ , it is defined that relay R forwards the symbol of W ₁ with the probability of p ₁ , forwards the symbol of W ₂ with the probability of p ₂ , and performs XOR forwarding with the probability of p ₃ , and p ₁ +p ₂ +p ₃ =1, then the joint decoding failure probability formula of destination D is: and

where, y _0,1 =y _0,2 =0, p' ₁ =p ₁ /1-p ₂ , p' ₂ =p ₂ /1-p ₁ , p' ₃ =p ₃ /1-p ₂ , p' ₄ =p ₃ /1-p ₁ .

Without loss of generality, assuming that the code lengths of the two sources and the deletion probabilities of the links are equal, we have: Assuming that the system symbols of source W ₁ and source W ₂ received by relay R are k ₁ and k ₂ respectively, the redundant code symbols are n ₁ and n ₂ respectively, and the deletion probability from relay R to destination D is 0 , the number of received redundant symbols m, then the ratio of coded symbols received by destination D is shown in Table 1.

The different symbol probabilities that the relay chooses to forward are shown in Table 2 below.

The distribution of the pseudo-random dispersion of the systematic Raptor code is as follows: Ω(x)=0.00976658x ¹ +0.459043x ² +0.210964x ³ +0.113393x ⁴ +0.111342x ¹⁰ +0.0798635x ¹¹ +0.0156279x ⁴⁰ .

by the formula and Assuming that the code length K=100, and the information source generates 140 coded symbols, the theoretical decoding performance of the DSRC scheme is shown in FIG. 5 .

And-or tree analysis is a classic tool for guiding and verifying the design of degree distribution. It can be seen from Figure 5 that the decoding asymptotic performance between the third method and the second method in this example is close, and the asymptotic performance of the third method is the best. good. In addition, for the Gaussian elimination decoding algorithm adopted by the ground station, the Gaussian decoding lower bounds of the three relay processing methods are simulated to give the same performance trend.

Next, specific data analysis is carried out through the simulation results, and the decoding failure rate (Decoding Failure Rate) is an important index to measure the performance of the distributed relay erasure code. To compare the decoding performance of the proposed DSRC with SLRC and IRDRS, 10 ⁴ Monte Carlo simulations are performed for each decoding redundant parameter point.

The performance simulation of the DSRC scheme is as follows:

First, the decoding performance of the DSRC scheme under different packet loss rates is simulated. Figure 6 and Figure 7 take the code length K=100 as an example, wherein Figure 6 shows P _wr,1 =P _wr,2 =0.1, the decoding performance of different P _rd is simulated; Figure 7 shows P _rd =0.1, and the simulation is different The decoding performance of P _wr,1 and P _wr,2 ; from Figure 6 and Figure 7, it can be seen that whether changing the packet loss rate P _wr from the source node to the relay node or changing the packet loss rate from the relay node to the destination D P _rd , the decoding success rate curves maintain a high degree of consistency. When the decoding redundancy is close to 0.1, the decoding failure probability is reduced to 10 ^-4 , which shows that the DSRC scheme has good robustness to the link packet loss rate. , which can adapt to different channel environments.

The comparison and simulation of the DSRC scheme with the existing SLRC and IRDRS schemes and the store-and-forward scheme (BF, Buffer and forward), the decoding failure probability performance curve is shown in Figure 8. It can be seen from Figure 8 that due to the application of the system Raptor code, The practicability of the DSRC scheme is greatly improved and the system overhead is reduced; and the XOR processing of the relay makes the DSRC scheme improve the decoding performance compared with BF.

In the performance simulation of the three methods in step S2 of the DSRC scheme in this example, the failure probability of relay decoding is set to 10 ^-4 , given by the formula It can be calculated that the number of encoded symbols to be received is 116, corresponding to link packet loss rates of 5% and 10%, and the source needs to send 122 and 128 encoded symbols. It is then set to generate 120/130 and 130/140 source coding symbols respectively, and the relay adopts three exclusive-or schemes of DSRC respectively, and the performance curve shown in Figure 9 is obtained. The first and second methods only use redundant coding symbols, and as the decoding redundancy increases to 240, the error platform is reached; while the third method uses both redundant coding symbols and system symbol, the decoding success rate can approach the performance bound of Gaussian decoding.

Based on the DTN asynchronous store-and-forward mechanism, this example designs a distributed system Raptor encoding scheme under the Y-shaped network for the scenario where multiple planetary surface probes transmit data to the ground through the same relay orbiter in the future. First, by optimizing the joint decoding performance of the destination, the complexity increase caused by Raptor codes in distributed systems is partially resolved. At the source end and relay R, the system structure codeword can greatly improve the data recovery efficiency. Theoretical analysis and simulation have verified that the DSRC scheme has a consistent decoding success rate when the deletion probability is less than 0.2, and it is consistent with existing Compared with other schemes of the technology, a decoding success rate of 99% is obtained when the redundancy reaches 5%.

In the future, we can further discuss the transmission scenario of unequal links from the Mars surface probe to the relay orbiter, and conduct distributed coding research on unequal errors for different business indicators and different code lengths to adapt to the diversified services of deep space communications in the future. need.

The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be assumed that the specific implementation of the present invention is limited to these descriptions. For those of ordinary skill in the technical field of the present invention, without departing from the concept of the present invention, some simple deduction or replacement can be made, which should be regarded as belonging to the protection scope of the present invention.

Claims (10)

1. a distributed system Raptor code transmission method facing deep space communication, it is characterized in that, comprising the following steps: Step S1, system Raptor codes the original information at the source end with code lengths K ₁ and K ₂ respectively and sends them to the same relay; Step S2, the relay R stores the encoded symbols from the source end in the buffer area _E1 and the buffer area _E2 respectively, and then uses the DSRC algorithm to process the data and then send it to the destination end; In step S3, the destination D decodes the received encoded symbols. 2. The distributed system Raptor code transmission method facing deep space communication according to claim 1, characterized in that, in the step S1, the system Raptor encoding process at the source end is after zero-filling by K source symbols C , through the intermediate generation matrix G constructed by the pseudo-random sequence, L intermediate symbols C' are generated, and satisfy the following formula: Among them, the intermediate generation matrix G includes S LDPC coded symbols G _LDPC , H Half coded symbols G _Half and K LT coded symbols G _LT , L=S+H+K, IS is a unit matrix of S rows and _S columns , 0 _{S × H} is the zero matrix of S rows and H columns, I _H is the unit matrix of H rows and H columns; S, L, H and K are all natural numbers, representing the rows and columns of the coding matrix; C' performs LT encoding to generate redundant coded packets, and the coded matrix is marked as R _LT ; then the original symbols and redundant coded symbols are sent in sequence. 3. the distributed system Raptor code transmission method facing deep space communication according to claim 1, characterized in that, in the step S1, the information source W ₁ and the information source W ₂ of the information source end carry out system Raptor encoding independently respectively Then send its original symbols and redundant encoded symbols to the relay R, and store them in the buffer area E ₁ and the buffer area E ₂ respectively; wherein, in step S2, the relay R passes through the buffer areas E _{1, w} and The buffer area E _{2, w} respectively buffers the original symbols of the information source W ₁ and the information source W ₂ , and buffers the redundant codes of the information source W ₁ and the information source W ₂ respectively through the buffer area E _{1, r} and the buffer area E _{2, r} symbol. 4. the distributed system Raptor code transmission method facing the deep space communication according to claim 3, it is characterized in that, in the described step S2, the coding symbols corresponding to the information source W ₁ and the information source W ₂ are respectively deleted with a probability of The erasure channels of P _wr1 and P _wr2 arrive at relay R. 5. The distributed system Raptor code transmission method facing deep space communication according to any one of claims 1 to 4, characterized in that, in the step S2, the relay R adopts the DSRC algorithm to process data and includes the following sub-steps : Step S201, forwarding the original symbols in the buffer area E _1,w or the buffer area E _2,w until all the buffered original symbols are forwarded; Step S202, realizing distributed encoding and forwarding through the relay R. 6. The distributed system Raptor code transmission method facing deep space communication according to claim 5, characterized in that, in the step S201, if the cache of buffer E _1,w or buffer E _2,w is non-empty , then randomly select an original symbol from the buffer area E _1,w or buffer area E _2,w to forward until all buffered original symbols are forwarded, then jump to step S202. 7. The distributed system Raptor code transmission method facing deep space communication according to claim 5, characterized in that, said step S202 realizes distributed code forwarding in any of the following three ways: Randomly from the buffer area Select a redundant coding symbol from E _{1, r} or buffer E _{2, r} to realize store-and-forward; randomly select a redundant coding symbol from buffer E _{1, r} and buffer E _{2, r} to realize XOR forwarding ; And, randomly select an original symbol from the buffer area E _1,w or the buffer area E _2,w , and randomly select a redundant coding symbol from the buffer area E _2,r or the buffer area E _1,r , and then realize XOR forward. 8. according to the described distributed system Raptor code transmission method facing deep space communication according to any one of claims 1 to 4, it is characterized in that, in the step S3, it is the RD segment chain of P _rd that the coding symbol is passed through deletion probability To reach the destination D, the process of joint matrix transformation is firstly simplified, and then Gaussian decoding is performed on the remaining stop set matrices that cannot be unitized. 9. the distributed system Raptor code transmission method facing deep space communication according to claim 8, is characterized in that, described step S3 comprises the following sub-steps: Step S301, initialize the state, and perform row elementary transformation and column elementary exchange on the decoding matrix; Step S302, realizing matrix simplification through row transformation and transformation elimination; In step S303, the unitization of the decoding matrix is realized, and the decoding ends. 10. the distributed system Raptor code transmission method facing deep space communication according to claim 9, it is characterized in that, in the described step S301, initialize state, carry out elementary conversion to sub _- block A1 and sub _- block A2 respectively First simplify the matrix with column elementary exchange, and then use the obtained I ₁ and I ₂ to perform elimination processing on R _LT , in which, the block A ₁ is the intermediate generator matrix G ₁ corresponding to the source W ₁ , and the block A ₂ is the signal The intermediate generator matrix G ₂ corresponding to the source W ₂ , I ₁ is the unit matrix obtained after the elementary transformation of the block A ₁ is simplified, I ₂ is the unit matrix obtained after the elementary transformation of the block A ₂ is simplified, and R _LT is A generation matrix of redundant symbols obtained by randomly XORing redundant symbols from two information sources; in the step S302, U ₁ and U ₂ are divided into upper and lower parts, and U _1,low is divided into upper and lower parts by row elementary transformation and U _2,low become an upper triangular matrix, and use the transformed U _1,low and U _2,low to perform row transformation and elimination on the block [0, B ₁ ,0, B ₂ ], and then select the appropriate Rows are supplemented to U _1,low and U _2,low to obtain C ₁ and C ₂ , where U ₁ is the non-zero number of L rows and unknown columns generated after the first-stage elementary transformation of block A ₁ Matrix; block U ₂ is the non-zero matrix with L rows and unknown number of columns generated after the first-stage elementary transformation of block A ₂ , and U _1,low is the upper triangular matrix after elementary transformation of U ₁ , U _2,low is the upper triangular matrix after the elementary transformation of U ₂ , the block [0, B ₁ , 0, B ₂ ] is the non-zero matrix after the elementary transformation of RLT, and the block C ₁ is the U _1. Perform elementary transformation to obtain the upper triangular matrix U _{1, the remaining submatrix after low} , block C ₂ is to carry out elementary transformation to U ₂ to obtain the upper triangular matrix U _{2, the remaining submatrix after low} ; in the step S303, rounding Remove the remaining M-2L rows, mark the blocks as A ₁₁ and A ₂₂ , and unitize them respectively, and the decoding ends, where M is the number of rows of the received joint decoding matrix, and L is the length K of the input source code The code lengths of the determined systematic codes, A ₁₁ and A ₂₂ are both simplified matrices of joint decoding.