Adaptive coding transmission method based on multi-task competition strategy in satellite networks
By adopting multi-task competition strategy and adaptive coding transmission method in satellite networks, combining bilateral search dynamic iteration algorithm and robust solitary wave distribution function, the problems of large transmission delay and link instability in satellite networks are solved, and small and medium-sized The efficient and reliable transfer of files improves the quality of service.
Patent Information
- Application Number
- CN202310036286.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-01-09
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2043-01-09
AI Technical Summary
The transmission delay, wide propagation loss and unstable links in satellite networks lead to low reliability and efficiency of data transmission. Especially under the rapidly changing network topology, it is difficult to ensure efficient transmission of small and medium-sized files.
Adaptive encoding transmission method based on multi-task competition strategy is adopted, and the encoded data packet size is adjusted through small file priority scheduling and bilateral search dynamic iterative algorithm, and the encoded packet is generated by combining the robust solitary wave distribution function, and fountain code and BP/GE decoding are used. The algorithm ensures reliable communication.
It effectively reduces the average waiting delay, improves service quality, ensures reliable and efficient transmission of multiple small and medium-sized files, and reduces the delay and bit error rate during the transmission process.
Smart Images

Figure CN116054917B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of satellite networks. Background Art
[0002] Satellite network is an important area for the development of the future communications industry. Satellites can provide a wider coverage and higher throughput than ground optical fiber and mobile network infrastructure. Compared with other communication methods, satellite communication has the advantages of long communication distance, large communication capacity, and strong seamless coverage. Researching satellite networks and promoting their development is of great significance to national development.
[0003] Satellite networks have disadvantages such as large transmission delay, wide propagation loss, and unstable links. The amount of data generated on satellites is large, the network topology of satellites changes frequently and rapidly, and the links are time-varying and unreliable. The application of coded transmission in satellite networks can effectively improve the stability of transmission and reduce delay. The present invention proposes a coded transmission method based on a multi-task competition strategy, which is used to reasonably sort the transmission of multiple small and medium-sized files in satellite communications and cope with the rapidly changing network topology, so as to ensure the reliable and efficient transmission of multiple small and medium-sized files. Summary of the invention
[0004] The present invention studies the reliability of data transmission over a satellite network, and proposes an adaptive coding transmission method based on a multi-task competition strategy for the transmission problem of multiple small and medium-sized files. Interruptible task scheduling is performed on multiple files to be transmitted, and a small file priority strategy is adopted for scheduling, and small files are encoded and transmitted preferentially, and long jobs and short jobs are fully weighed, taking care of short jobs while not making the long jobs wait too long, effectively reducing the average waiting delay and improving the service quality. The size of the coded data packet is adaptively adjusted according to the size of the transmitted file and the network environment, and a bilateral search dynamic iterative algorithm is proposed to find the optimal size and grouping of the coded packet, avoiding the need for frequent feedback during the transmission process, ensuring reliable communication, and reducing the receiving delay.
[0005] (1) Scheduling interruptible data transfer for file tasks
[0006] Sort the files waiting to be transmitted by the sender by size, and give priority to transmitting small files. Each file has its own maximum waiting time limit D. If the file currently being transmitted does not exceed its maximum waiting time limit, its transmission can be interrupted. Otherwise, the current file will continue to be transmitted first. If a new file is added to the transmission queue of the sender during the transmission process, the size of the new file is smaller than the remaining size of the current transmission file, and the current transmission meets the interruption condition, the new file will be transmitted first. The average waiting delay of the mth task The minimization of can be modeled as:
[0007]
[0008]
[0009] Among them, t j represents the current total waiting time of the jth file task, D j Indicates the maximum waiting time limit for the j-th file task, and N is the total number of tasks.
[0010] (2) According to the satellite network status, the bilateral search dynamic iterative algorithm is used to solve the number of file segmentation groups.
[0011] In the process of data transmission in satellite network, there will be delay, including transmission delay, link delay and processing delay. The transmission delay Tt is expressed as Where Q is the size of a single coded packet, r is the data transmission rate between the sender and the receiver. The processing delay is the time required for encoding and decoding, and the processing delay T p Expressed as formula T p =CM, where M is the data file size and C is the computing power of data encoding and decoding, expressed in ms / MB, i.e., the time required to process 1MB of data.
[0012] The source file to be transmitted is divided into n information packets. The encoder encodes n information packets to generate coded packets. The coded packets are transmitted through the transmission channel. The receiving end can decode and obtain the original data after receiving enough coded packets. The expected value of the number of coded packets is calculated by the number of coded packets expected to be received by the receiving end and the number of coded packets finally received. A bilateral search dynamic iteration algorithm is proposed to solve the optimal number of information packets. The algorithm adopts a heuristic algorithm, which can dynamically iterate from both sides and obtain the final optimal value by comparing the results of the iterations on both sides.
[0013] (3) Adaptive multi-mode fusion data encoding method
[0014] The classic LT code suitable for a large RTT environment is adopted, and the robust solitary wave distribution function is applied to generate the coding packet degree. The robust solitary wave distribution function is shown as μ(i).
[0015]
[0016] Where θ = ∑ i (ρ(i)+τ(i)), ρ(i) is the ideal solitary wave distribution function, ρ(i) represents the probability that the degree of the output symbol node is i, and the ideal solitary wave distribution function can be expressed as follows.
[0017]
[0018] Where n is set to the number of segmentation groups calculated according to the satellite network status. The robust solitary wave distribution adds τ(i) to the design of the ideal solitary wave distribution, which can increase the number of packets with degree 1 in the decoding process to multiple. The expression of τ(i) is shown in the following formula.
[0019]
[0020] Among them, δ is the maximum failure probability of decoding, R is the number of coding groups with degree 1, and R can be expressed as:
[0021]
[0022] Where c is a constant and is greater than 0. Randomly select a degree value i from the robust solitary wave distribution, calculate the XOR sum of i different data packets, and generate a coded packet. Repeat this step to obtain multiple coded groups.
[0023] (4) Dynamically coded packet receiving and feedback method
[0024] After receiving the coded packet, the receiver uses BP (backward propagation) and GE (Gaussian elimination) decoding algorithms, and replies with a confirmation message after all decoding is completed. Let the message matrix after the original data is fragmented be S, and the encoding process is the multiplication of the message matrix and the generation matrix G. Let the matrix after encoding be E, then the encoding process can be expressed as E=SG. G is a K*L matrix randomly generated according to the degree distribution. The number of 1s in each column of G represents the degree, and the position represents the sequence number of the source data packet involved in the encoding. The encoding process can be expressed as follows:
[0025]
[0026] where e j represents the jth coded packet, s j Represents the jth original data packet. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION
[0028] In the satellite file transmission system, data is transmitted in units of files. The present invention adopts the form of fountain code to transmit data. First, the original file needs to be segmented to obtain information packets. Considering the instability of the satellite transmission channel, the present invention adopts a typical fountain code-LT code, which is a practical fountain code in the 3GPP standard and is suitable for environments with large RTT. Since the satellite transmission channel is unstable and unreliable, there is a possibility of bit error in each bit during the data transmission process. It is also necessary to add check information in the encoding packet to determine whether a single data packet is transmitted correctly. The specific implementation scheme of the present invention is as follows:
[0029] (1) Real-time and interruptible adjustment of the order of the transmission sequence
[0030] The present invention simulates satellite network transmission and transmits multiple file tasks between two satellites. The probability of task transmission in each cycle follows Poisson distribution, and the size of random tasks is 8.18MB, 131MB and 51.9MB. 100 cycles are simulated, and the average throughput is calculated by the transmission file size and transmission time.
[0031] Adjust the transmission order in real time. Specifically, if the file currently being transmitted is a large file of 131MB, after the small file of 8.18MB arrives, if the waiting time of the current transmission task does not exceed the upper limit of the waiting time, compare the size of the current transmission file with the new file, and give priority to the encoding transmission of the small file, that is, interrupt the transmission of the 131MB file, and give priority to the newly arrived 8.18MB small file; if the waiting time of the current transmission task exceeds the upper limit of the waiting time, the transmission of the current file cannot be interrupted, and the current file is given priority, that is, the large file of 131MB continues to be transmitted to ensure that the file is transmitted to the receiving end within its acceptable waiting time. In addition, during the transmission process, each file has an ID. After the encoded packet arrives at the receiving end, the receiver can decode it in multiple threads at the same time according to the ID of the received packet.
[0032] (2) Calculate the optimal number of information packets using the bilateral iterative search algorithm based on the satellite network status
[0033] The transmission delay is Where M is the data file size, which can be 8.18MB, 131MB or 51.9MB, n is the number of packets after the file is segmented, a is the encoding parameter, a is set to 0.125, b is the checksum, and each packet uses 1 bit to mark whether it participates in the encoding, r is the data transmission rate between the sender and the receiver, and the link transmission speed is set to a Gaussian random number with an average value of 1MB / s and a standard deviation of 0.05MB / s. The processing delay is T p =CM, C is the computing power of data encoding and decoding.
[0034] Encoded transmission
[0035] As shown in the figure above, the file is transmitted between two satellites. The source file is divided into n information packets. The encoder encodes the n information packets to generate E(n s ) code packets, E(n s ) is the expected value of the number of coded packets, and the number of coded packets expected to be received by the receiving end is equal to the number of coded packets finally received, that is, E(n s )(1-P)=n(1+ε), we can calculate the expected value of the number of coded packets E(n s), where P is the probability of failure of transmission of a single data packet and ε is redundancy.
[0036] The problem of minimizing transmission delay can be modeled as follows:
[0037] minE(T),
[0038]
[0039]
[0040]
[0041] Among them, E(T) is the expectation of the total transmission delay, p is the bit error rate, and the link bit error rate value interval is set to 10 -10 -10 -6 , T t is the transmission delay of a single data packet, and r is the data transmission rate. According to the above formula, the approximately optimal number of information packets n is calculated.
[0042] The final expression of E(T) can be organized as follows:
[0043]
[0044] Since the above formula is difficult to solve, the approximate formula without rounding upward is solved, and the approximate formula is as follows:
[0045]
[0046] Analyze the approximate formula, the numerator is a quadratic function, and a>0, The minimum value is taken at n = 0. Denominator In, due to 1-p<1, when the inequality is rounded, It takes the maximum value when . From this, we can deduce that the approximate formula has a lower bound in the interval (0, +∞). When n approaches 0 from the positive direction, the approximate formula tends to infinity. Since the formula is in the interval Monotonically increasing, when Therefore, the approximate formula can be derived in Get the minimum value.
[0047] By taking the derivative of the approximate formula and setting the derivative equal to 0, we can get the following formula:
[0048] -8ln(1-P)a 2 n 4 +(2a-8ln(1-P)ab)n 3 +bn 2+8ln(1-P)bMn+8ln(1-P)M 2 =0
[0049] Since the number of information packets n is an integer, it is difficult to directly solve the value of n. Therefore, a heuristic algorithm is used to solve it. A bilateral search dynamic iterative algorithm is proposed to iterate from both sides. In the above formula, the constant term is less than 0 and the coefficient of the quartic term is greater than 0, so the equation has at most four roots. According to Vieta's theorem, it can be inferred that the four roots have two distributions, three positive roots and one negative root or three negative roots and one positive root. When n = 1, the value of the above formula is less than 0 and the derivative is less than 0. Compare the gradient descent result on the right side of n = 1 with The results of the gradient descent on the left are compared. If the n value of the gradient descent on the left is less than 0, the result of the gradient descent on the right is taken; otherwise, the gradient descent results on both sides are compared and the smaller value is taken. The value obtained is the optimal n value.
[0050] (3) Encode the file based on the optimal number of packets calculated based on the adaptive network status
[0051] The encoding and transmission are performed according to the encoding parameter a, the file size M, the redundancy ε and the optimal number of information packets calculated in step (2). The classic LT code suitable for a large RTT environment is used, and the robust solitary wave distribution function is applied. Generate a coded packet number i, randomly select the original data packet number, perform XOR operation on i information packets to obtain a coded packet, and repeat this step to obtain multiple coded packets.
[0052] (4) Send the encoded packet to the receiving end
[0053] The coded packets generated in step (3) are continuously transmitted from the transmitter to the receiver via the satellite link, and the link bit error rate is set to 10 -1 0-10 -6 It can be calculated that the greater the bit error rate, the greater the impact of the adaptive coding transmission algorithm on delay reduction.
[0054] (5) The receiving end decodes the received adaptive coding packet and obtains the source data
[0055] The BP and GE decoding algorithms are used. The encoding process can be expressed as E=SG, where E is the encoded matrix, S is the message matrix after the original file is segmented, and G is the generator matrix. The encoded matrix and the message matrix can be expressed as E(n s ) is the number of coded packets calculated in step (2), and k is the length of the LT code. When the rank of G is equal to k, the system of equations can be solved and the source message can be decoded; when the rank of G is less than k, the receiver needs to continue to receive data packets until the system of equations can be solved.
[0056] The encoding process can be expressed as follows:
[0057]
[0058] where e j represents the jth coded packet, s j represents the jth original data packet. The process of solving the above equation group is carried out through GE decoding. The decoding process is to find the number of 1s each time. j Decoding is performed, and enough coded packets are received to make the generator matrix full rank, so as to reversely decode the original data packet and return a confirmation message.
Claims
1. An adaptive coding transmission method based on a multi-task competition strategy in a satellite network, characterized in that The following steps are involved: (1) Scheduling interruptible data transfer for file tasks Sort the files waiting to be transmitted by the sender in order of size, and give priority to transmitting small files; each file has its own maximum waiting time limit D. If the file currently being transmitted does not exceed its maximum waiting time limit, its transmission will be interrupted, otherwise, the current file will continue to be transmitted first; If a new file is added to the transmission queue of the sender during the transmission process, the size of the new file is smaller than the remaining size of the current transmission file, and the current transmission meets the interruption condition, the new file will be transmitted first; the average waiting delay of the mth task The minimization model is: Among them, t j represents the current total waiting time of the jth file task, D j Indicates the maximum waiting time limit of the j-th file task, and N is the total number of tasks; (2) According to the satellite network status, the bilateral search dynamic iterative algorithm is used to solve the number of file segmentation groups. In the process of data transmission in satellite network, time delay will occur, including transmission delay, link delay and processing delay. t Expressed as a formula Where Q is the size of a single coded packet, r is the data transmission rate between the sender and the receiver; the processing delay is the time required for encoding and decoding, and the processing delay T p Expressed as formula T p =CM, where M is the size of the data file and C is the computing power of data encoding and decoding, in ms / MB, i.e., the time required to process 1MB of data; The source file to be transmitted is divided into n information packets. The encoder encodes the n information packets to generate coded packets. The coded packets are transmitted through the transmission channel. After receiving enough coded packets, the receiving end decodes them to obtain the original data. The expected value of the number of coded packets is calculated by the number of coded packets expected to be received by the receiving end and the number of coded packets finally received. (3) Adaptive multi-mode fusion data encoding method The LT code suitable for the environment with large RTT is adopted, and the robust solitary wave distribution function is applied to generate the coding packet degree. The robust solitary wave distribution function is shown as μ(i); Where θ = ∑ i (ρ(i)+τ(i)), ρ(i) is the ideal solitary wave distribution function, ρ(i) represents the probability that the degree of the output symbol node is i, and the ideal solitary wave distribution function is expressed as follows; Where n is set to the number of segmentation groups calculated according to the satellite network status. The robust solitary wave distribution adds τ(i) to the design of the ideal solitary wave distribution, which can increase the number of groups with degree 1 in the decoding process to multiple. The expression of τ(i) is shown in the following formula: Among them, δ is the maximum failure probability of decoding, R is the number of coding groups with degree 1, and R is expressed as: Where c is a constant and is greater than 0; a degree value i is randomly selected from the robust solitary wave distribution, and the i different data packets are XORed and summed to generate a coded packet; this step is repeated continuously to obtain multiple coded groups; (4) Dynamically coded packet receiving and feedback method After receiving the coded packet, the receiver uses the back propagation BP and Gaussian elimination GE decoding algorithms, and replies with the feedback confirmation message after all decoding is completed; let the message matrix after the original data is fragmented be S, and the encoding process is the multiplication of the message matrix and the generation matrix G; let the matrix after encoding be E, then the encoding process is expressed as E=SG; G is a K*L matrix randomly generated according to the degree distribution, the number of 1s in each column of G represents the degree, and the position represents the sequence number of the source data packet involved in the encoding; the encoding process is expressed as the following formula: where e j represents the jth coded packet, s j represents the jth original data packet; Among them, the problem of minimizing transmission delay is modeled as the following expression: min E(T), Among them, E(T) is the expectation of the total transmission delay, p is the bit error rate, and the link bit error rate value interval is set to 10 -10 -10 -6 , T t is the transmission delay of a single data packet, r is the data transmission rate; the approximately optimal number of information packets n is calculated according to the above formula; The final expression of E(T) is organized as follows: Since the above formula is difficult to solve, the approximate formula without rounding upward is solved, and the approximate formula is as follows: Analyze the approximate formula, the numerator is a quadratic function, and a>0, Take the minimum value at n = 0; the denominator In, due to 1-p<1, when the inequality is rounded, It takes the maximum value when ; from this, it is deduced that the approximate formula has a lower bound in the interval (0, +∞); when n approaches 0 from the positive direction, the approximate formula tends to infinity; since the formula is in the interval Monotonically increasing, when Therefore, the approximate formula is derived in Get the minimum value; Derivative the approximate formula and set the derivative equal to 0 to obtain the following formula: -8ln(1-P)a 2 n 4 +(2a-8ln(1-P)ab)n 3 +bn 2 +8ln(1-P)bMn+8ln(1-P)M 2 =0 A bilateral search dynamic iteration algorithm is proposed, which iterates from both sides. In the above formula, the constant term is less than 0 and the coefficient of the quartic term is greater than 0, so the equation has at most four roots. According to Vieta's theorem, it is inferred that there are two distributions of the four roots, three positive roots and one negative root or three negative roots and one positive root. When n = 1, the value of the above formula is less than 0 and the derivative is less than 0. The gradient descent result on the right side of n = 1 is compared with The results of the gradient descent on the left are compared. If the n value of the gradient descent on the left is less than 0, the result of the gradient descent on the right is taken; otherwise, the gradient descent results on both sides are compared and the smaller value is taken; the value obtained is the optimal n value.
Citation Information
Patent Citations
Power distribution method based on MIMO-NOMA short packet transmission
CN110492915A
Communication method and device based on Internet of Vehicles, computer equipment and storage medium
CN112333670A
Probabilistic feedback and staged fountain code transmission method in self-organizing network
CN112887056A
Reliable data transmission method based on reliable UDP and fountain code in aeronautical ad hoc networks
US20190190542A1
