CN116866246A - A path construction method for satellite data transmission - Google Patents

Abstract

The invention provides a path construction method for satellite data transmission. The method includes: after the data request direction sends a data transmission request to the satellite, the satellite labels all paths and sends several data packets to each path; using The Gilbert model calculates the packet loss expectation of each path and draws the path topology diagram; based on the path topology diagram, the Markov decision method is used to dynamically select through the four processes of establishing a state set, obtaining an action set, calculating the state transition probability, and obtaining feedback The best path selection plan is to establish a fluid neural network model based on the best path selection plan, set several data nodes on the fluid neural network model, and select the path channel with the largest flow to connect the data nodes to obtain the best path.

Description

A path construction method for satellite data transmission

Technical field

The present invention relates to the field of satellite networks, and in particular to a path construction method for satellite data transmission.

Background technique

With the continuous development of the Internet, the amount of information transmission generated every day is also increasing, and satellites are moving at high speed. Therefore, the nodes of the satellite are also moving at high speed, and the links are frequently switched, causing the transmission path to fail;

Existing SDN technical solutions are mainly optimized for terrestrial networks. The optimization effect is not ideal for satellites, a network environment with a large space and time. Moreover, satellites have more and more network nodes and the link structure becomes more and more complex, making satellites During data transmission, the round-trip delay of the path increases, which has also become a major constraint on the efficiency of satellite data transmission. Therefore, this method provides a path construction method for satellite data transmission.

Contents of the invention

A path construction method for satellite data transmission, including:

Step 1: After receiving the data transmission request, the satellite numbers all paths and sends several data packets to each path;

Step 2: Use the Gilbert model to calculate the packet loss expectation for each path;

Step 3: Draw the path topology map;

Step 4: Use Markov decision-making method to dynamically select the best path selection plan;

Step 5: Establish a fluid neural network model based on the optimal path selection plan;

Step 6: Search for the path with the largest flow in the fluid neural network model to obtain the best path.

Further, the satellite's path processing process includes:

After receiving the data transmission request, the satellite numbers all the paths, sends several data packets to each path at the same time, and counts the number of data packets successfully transmitted to the data requester on each path, and then counts the packet loss of each path. quantity.

Further, the calculation process of packet loss expectation for each path includes:

Perform three-layer loops on each path to calculate the path packet loss rate. The number of loops in the first layer is the total number of paths, the number of loops in the second layer is the number of data packets sent by the satellite through the path, and the number of loops in the third layer is the path. The number of lost packets;

Calculate the probability of losing several data packets when the path allocation predicts data packets in each cycle, and then calculate the mathematical expectation of losing data packets on this path.

Further, the process of drawing the path topology map includes:

Set a packet loss threshold and compare it with the mathematical expectation of packet loss on a certain path;

If the packet loss threshold is less than or equal to the mathematical expectation, set the label "available" for the path and add the path to the path topology map;

If the packet loss threshold is greater than the mathematical expectation, the label is set to "unusable" and the path is temporarily hidden;

Repeat the above operations until all paths that meet the conditions are included in the path topology map.

Further, the calculation process of the Markov decision method includes:

The Markov decision-making method includes four processes: establishing a state set, obtaining an action set, calculating the state transition probability, and obtaining feedback;

Obtain the total number of available paths, round-trip delay, packet loss rate and maximum window based on the path topology map;

Establish a state set, record the status of each available path, calculate and label the window size in the congestion state, and integrate it into a state set;

Obtain an action set. The action elements in the action set represent all optional path selection decisions. The optional path selection decisions indicate that the parallel transmission throughput is guaranteed to be less than or equal to the maximum single path throughput to prevent instantaneous throughput of path data transmission. Path selection decisions that are too large cause path congestion;

Calculate the state transition probability, based on the joint congestion mechanism, detect whether the path accepts or loses data packets. If the path receives data packets, the path window will be increased. If the data packets are lost, the path window will be reduced. After executing the path selection action decision, the path passes through After multiple times of congestion control, after the absolute value of the window changes to 1, the state transition probability is calculated based on the window size and the number of data packets received or lost;

Obtain feedback, which represents the instantaneous data throughput of the path after executing the path selection action decision.

Further, the dynamic selection process of the best path selection plan includes:

Set instantaneous data throughput threshold and compare feedback;

If the feedback is less than the instantaneous data throughput threshold, the path selection decision will be deleted from the action set. If the feedback is greater than or equal to the instantaneous data throughput threshold, the path selection decision will be left in the action set;

Repeat the above operation to delete all path selection decisions whose action centralized feedback is less than the instantaneous data throughput threshold, and integrate the remaining path selection decisions to obtain the best path selection plan.

Further, the establishment process of the fluid neural network model includes:

The fluid neural network model is a neural network model with fluid characteristics. In the fluid neural network, the probability of a single fluid molecule flowing through a certain channel is related to the flow rate of the channel;

After receiving the confirmation response from the data requester, the satellite constructs a fluid neural network model based on the path in the optimal path selection plan, and constructs several data nodes in the fluid neural network model.

Further, the process of obtaining the best path includes:

Analogous to the probability of fluid molecules flowing through a channel in a fluid neural network and the channel flow rate, during the data transmission process, the probability of data passing through a path is related to the flow rate of the path;

Then select the channel with the largest flow between one data node and its next data node to connect, and repeat the above operations until all data nodes in the fluid neural network are connected, and then the best path is obtained.

Compared with the prior art, the beneficial effects of the present invention are:

This method uses the Markov decision method to establish the path selection decision, fully considers the impact of the round-trip delay and packet loss rate during data transmission on the path throughput, and can effectively balance the flow, providing a basis for selecting the optimal path in the fluid neural network model. The best path is guaranteed.

Description of the drawings

Figure 1 is a flow chart of the present invention.

Implementation

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.

As shown in Figure 1, an embodiment of the present invention provides a path construction method for satellite data transmission. The method includes the following steps:

Step 1: The data requesting direction sends a data transmission request to the satellite. After receiving the data transmission request, the satellite numbers all paths, such as A ₁ , A ₂ ,..., A _i , and sends N data packets to each path simultaneously. , by counting the number of data packets successfully transmitted to the data requester on each path, and then counting the number of packet losses on each path, where N=1, 2, 3...

Step 2: Use the Gilbert model to calculate the packet loss expectation P for each path;

Specifically, based on the operation in step 1, the number of lost packets for each path is predicted to be 1, 2, 3...t, where t=1, 2, 3..., and t≤N;

Perform three-layer loops on each path to calculate the path packet loss rate P. The number of loops in the first layer is the total number of paths Num. The number of loops in the second layer is the number N of data packets sent by the satellite through the path. The number of loops in the third layer is is the number of lost packets T on this path;

Further, calculate the probability of losing a data packet when N data packets are allocated to each cycle path, and then calculate the mathematical expectation M of losing data packets on this path. The calculation formula is:

Among them, M _i represents the mathematical expectation of losing data packets when allocating N data packets for the path number A _i , t _i represents the number of times that i data packets are lost in the cycle, T _i represents the current number of cycles, and a _i represents The number of lost packets in the i-th cycle.

Step 3: Draw the path topology map;

Set the packet loss threshold p, the mathematical expectation M for packet loss on a certain path;

If M ≤ p, set the label "available" for the path and add the path to the path topology map;

If M>p, set the label "not available" for the label and temporarily hide the path;

Merge all paths with the "available" label into the path topology map to obtain the complete path topology map.

Step 4: Use Markov decision-making method to dynamically select the best path selection plan;

The Markov decision-making method dynamically selects paths according to the size of each path window, and selects different paths for transmission in real time based on the optimal decision during data transmission. This method can optimize the balanced flow and will not affect the data during data transmission. Robustness of transmission pairs and improved throughput when doing parallel robustness;

Determining the best path selection decision in the Markov decision-making method requires four processes: state set, action set, state transition probability, and feedback.

Create a state set;

The total number of available paths N is obtained based on the path topology map, where the round-trip delay, packet loss rate and maximum window of each available path are recorded as T _n , P _n , W _n respectively, where n=1,2,3...;

Each available path has Markov properties, and its window size at the next moment is only related to the window size at the current moment. The status of each available path is recorded, and the window size in the congestion state is calculated and labeled, such as S ₁ , S ₂ ..., and integrate them into a state set S:

S={（S ₁ , S _2, ,……）}

Among them, S _n =βW _n , β is the congestion coefficient, and β∈(0,1).

Get action set;

The action elements in the action set represent different path selection decisions. For example, there are 2 ^N -1 decisions for N paths;

Specifically, the path selection in the Markov decision-making method may switch from a small number of paths to a multi-path number. In this case, the important goal of joint congestion control is to ensure that the satellite achieves TCP fairness in parallel transmission during the data transmission process. sex;

Furthermore, in the path selection decision selected under the circumstances, it is necessary to ensure that the parallel transmission throughput is less than or equal to the maximum single path throughput to prevent path congestion caused by excessive instantaneous throughput of path data transmission;

Further, set the action set G, and G = {g ₁ , g ₂ ,..., g ₂ ^N _-1 }, where g _m expresses the path selection action decision, where m = 1, 2, 3....

Calculate state transition probability;

Based on the joint congestion mechanism, the congestion control process when using action decisions includes:

If path A _i accepts a data packet, the congestion window W _n of path A _i will increase by 1/βW _n ;

If packet loss occurs on path A _i , the congestion window W _n of path A _i will be reduced to βW _n /2;

Furthermore, after the path selection action decision g _m is executed and the absolute value of the window change of path A _i is 1 after multiple congestion controls, the state transition probability H can be expressed as:

Obtain feedback, which represents the instantaneous data throughput of the path after executing the path selection action decision;

When multiple paths transmit data in parallel, the round-trip delays of each path are different, and the time it takes for each data packet to arrive at the data requester is also different. During data transmission, since the data requester needs to wait for the arrival of the data packet, the round-trip Delay will also affect the throughput of data transmission. If feedback R is set, the calculation formula of feedback R is:

Further, set the instantaneous data throughput threshold r, and compare the instantaneous data throughput after the path selection action decision is executed with the instantaneous data throughput threshold r according to the feedback R;

If the feedback R is greater than or equal to the instantaneous data throughput threshold r, the path selection decision will be left in the action set;

If the feedback R is less than the instantaneous data throughput threshold r, the path selection decision will be deleted from the action set;

Repeat the above steps in sequence until all path selection decisions whose action centralized feedback R is less than the instantaneous data throughput threshold r are deleted;

The remaining path selection decisions are integrated to obtain the best path selection solution.

Step 5: The satellite transmits the best path selection plan to the data requester. After confirmation by the data requester, the satellite deploys the fluid neural network model of the data transmission path according to the best path selection plan and the data requester;

Specifically, after receiving the confirmation response from the data requester, the satellite constructs a fluid neural network model based on the path in the path construction plan;

The fluid neural network model is a neural network model with fluid characteristics. Assume that each neuron of the fluid neural network is a container filled with fluid, then in the fluid neural network containing x containers, the yth container The dynamic characteristics are as follows:

Among them, w _yl represents the communication ability of the channel connecting containers y and l, I _y represents the input of the container, s _y represents the height of the liquid in the container, and u _y represents the volume of the y-th container.

In the fluid neural network, the probability of a single fluid molecule flowing through a certain channel is related to the flow rate of the channel;

By analogy, during data transmission, the probability of data passing through a path is related to the traffic of the path;

Further, several data nodes are constructed in the fluid neural network model, and the satellite's transmission starting point and the data receiver's receiving end point are labeled λ and μ, and the intermediate data nodes are labeled z ₁ , z ₂ , ....

Step 6: Search for the path with the largest flow in the fluid neural network model to obtain the best path;

Starting from the transmission starting point λ, search for the path with the largest traffic to the next path node z ₁ , and then search for the path with the largest traffic from z ₁ to z _2. Repeat the above operation until the end point μ is accepted;

λ→z ₁ →z ₂ ……→μ in the fluid neural network is the optimal path for satellite data transmission.

The above embodiments are only used to illustrate the technical methods of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical methods of the present invention can be modified or equivalently substituted. without departing from the spirit and scope of the technical method of the present invention.

Claims (8)

1. A path construction method for satellite data transmission, comprising:

step one, after receiving a data transmission request, a satellite numbers all paths and sends a plurality of data packets to each path;

calculating the data packet loss expectation of each path by adopting a Gilbert model;

drawing a path topological graph;

step four, dynamically selecting an optimal path selection scheme by adopting a Markov decision method;

establishing a fluid neural network model based on an optimal path selection scheme;

and step six, searching a path with the maximum flow in the fluid neural network model to obtain an optimal path.

2. A method of path construction for satellite data transmission according to claim 1, wherein the satellite-to-path processing comprises:

after receiving the data transmission request, the satellite numbers all paths, simultaneously sends a plurality of data packets to each path, and counts the number of the data packets successfully transmitted to the data requesting party by each path, thereby counting the number of lost packets of each path.

3. A method of path construction for satellite data transmission according to claim 2, wherein the calculation of packet loss expectations for each path comprises:

respectively carrying out three-layer circulation on each path so as to calculate the packet loss rate of the path, wherein the circulation times of the first layer are the total number of paths, the circulation times of the second layer are the number of data packets sent by the satellite through the path, and the circulation times of the third layer are the packet loss number of the path;

and calculating the probability of losing a plurality of data packets when the path of each cycle distributes the predicted data packets, and further calculating the mathematical expectation of the lost data packets of the path.

4. A path construction method for satellite data transmission according to claim 3, wherein the drawing process of the path topology map comprises:

setting a data packet loss threshold value, and comparing the data packet loss threshold value with mathematical expectations of a data packet lost by a certain path;

if the data packet loss threshold is less than or equal to the mathematical expectation, setting a label 'usable' for the path, and dividing the path into a path topological graph;

if the data packet loss threshold is greater than the mathematical expectation, setting a label 'unusable' for the label, and temporarily hiding the path;

repeating the above operation until all the paths meeting the conditions are all drawn into the path topology diagram.

5. A path construction method for satellite data transmission according to claim 4, wherein the calculation process of the markov decision method comprises:

the markov decision method includes four processes: establishing a state set, obtaining an action set, calculating state transition probability and obtaining feedback;

acquiring the total number of available paths, round trip delay, packet loss rate and maximum window based on a path topological graph;

establishing a state set, recording the states of all available paths, calculating and marking the size of a window in a congestion state, and integrating the window into the state set;

acquiring an action set, wherein action elements in the action set represent all selectable path selection decisions, and the selectable path selection decisions represent path selection decisions which ensure that the parallel transmission throughput is smaller than or equal to the maximum single path throughput so as to prevent path congestion caused by overlarge instantaneous throughput of path data transmission;

calculating state transition probability, detecting whether a path receives or loses a data packet based on a joint congestion mechanism, increasing a path window if the path receives the data packet, reducing the path window if the path receives the data packet, executing path selection action decision, and calculating the state transition probability according to the window size and the data packet receiving or losing quantity after the absolute value of window change is 1 after the path passes through congestion control for many times;

feedback is obtained, which represents the instantaneous throughput of data of the path after the path selection action decision has been performed.

6. The path construction method for satellite data transmission according to claim 5, wherein the dynamic selection process of the optimal path selection scheme comprises:

setting a data instant throughput threshold value and comparing feedback;

if the feedback is less than the data instant throughput threshold, deleting the path selection decision in the action set, and if the feedback is greater than or equal to the data instant throughput threshold, leaving the path selection decision on the action set;

repeating the above operations, deleting all path selection decisions fed back in the action set, wherein the feedback is smaller than the data instant throughput threshold, and integrating the rest path selection decisions to obtain the optimal path selection scheme.

7. The path construction method for satellite data transmission according to claim 6, wherein the fluid neural network model building process comprises:

the fluid neural network model is a neural network model with fluid characteristics, and in the fluid neural network, the probability of a single fluid molecule flowing through a certain channel is related to the flow rate of the channel;

after receiving the confirmation response of the data requesting party, the satellite builds a fluid neural network model according to the path in the optimal path selection scheme, and builds a plurality of data nodes in the fluid neural network model.

8. The path construction method for satellite data transmission according to claim 7, wherein the obtaining of the optimal path comprises:

the probability of fluid molecules flowing through a channel in the analog fluid neural network is related to the flow rate of the channel, and the probability of data passing through a path is related to the flow rate of the path in the data transmission process;

and then selecting a channel with the largest flow rate from one data node to the next data node for connection, and repeating the operation until all the data nodes in the fluid neural network are connected, so as to obtain an optimal path.