A method for optimizing satellite network transmission based on business forecasting

Abstract

The present invention is applicable to the field of satellite network communication technology, and provides a method for optimizing satellite network transmission based on business prediction for integrated communication and remote sensing, including: constructing a system model including satellites, user equipment and ground stations; proposing two remote sensing information transmission modes, the first mode is applicable to the situation where users have high requirements for remote sensing information and need to perform attitude maneuvers; the second mode is that the integrated communication and remote sensing satellite uses a flat scanning method to collect remote sensing information and assign time and space attributes, and the user equipment uploads the required remote sensing information and queries it in the ground station. Business prediction and inter-satellite routing planning are performed based on these two businesses. According to the business prediction results, an improved ant colony algorithm is used to obtain a path with the lowest transmission delay, a transmission bandwidth that meets the requirements and a balanced link load. The present invention can meet the communication needs of different user equipment, and can effectively predict business traffic, pre-allocate resources, reduce transmission delay, and improve the overall performance of the communication system.

Description

Technical Field

[0001] The invention belongs to the technical field of satellite network communications, and in particular relates to a communication and remote control integrated satellite network transmission optimization method based on business prediction. Background Art

[0002] Space remote sensing has entered a new stage of high temporal resolution, high spatial resolution, and high spectral resolution. People's demand for remote sensing services is constantly increasing. Coupled with the rapid development of satellite communication technology, the model of providing services by a single remote sensing satellite is gradually broken. Multiple low-orbit remote sensing satellites form a space information network, which can provide integrated and networked services for in-orbit acquisition, processing, and distribution of remote sensing data. The low-orbit communication and remote sensing integrated satellite network can not only complete high-coverage, all-weather, and near-real-time earth observation tasks, but also can get rid of the limitations of satellite transit time with the help of onboard routers and inter-satellite links, and realize real-time data downlink, which is very important in terms of business prospects and national strategic status. It is of great significance to study how to transmit remote sensing data back efficiently and reliably. For this reason, a communication and remote sensing integrated satellite network transmission optimization method based on business prediction is proposed. Summary of the invention

[0003] The purpose of the present invention is to provide a satellite network transmission optimization method based on business prediction for integrated communication and remote sensing, aiming to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method for optimizing satellite network transmission based on business prediction for integrated communication and remote sensing includes the following steps:

[0006] Step S1: construct a system model including satellites, user equipment and ground stations;

[0007] Step S2: Propose two methods of collecting remote sensing data by integrated satellite communication and remote sensing, and select one method of collecting remote sensing data;

[0008] Step S3: predicting network traffic data within a future time interval through historical remote sensing data;

[0009] Step S4: Calculate the routing table using the improved ant colony algorithm according to the predicted service traffic.

[0010] Furthermore, in step S2, the two communication and remote sensing integrated satellites collect remote sensing data in the following manner:

[0011] The first method: the user device uploads the location to be photographed, and the ground station uses the star selection algorithm to determine the shooting star and the backup shooting star based on the satellite ephemeris information, and uploads the shooting information to enable the shooting star to perform the operation. When the shooting star performs attitude maneuvers for shooting, it uploads the execution of the shooting task to the ground station; after the shooting star finishes the operation, it sends information to the ground station, and then the ground station uses the satellite routing algorithm, with the shooting star as the source node and the satellite above the user device as the destination node. According to the satellite node and link information, the routing algorithm is used to calculate the routing table from the source node to the destination node and transmit it to the satellite. After receiving the routing table, the satellite transmits remote sensing data;

[0012] The second type: the user equipment transmits the time information and space information of the required remote sensing information to the ground station. The satellite uses flat scanning to collect remote sensing data. Each satellite assigns time attributes and space attributes to the collected remote sensing information. Then each satellite transmits the time attributes and space attributes of its own remote sensing information block back to the ground station to form a remote sensing information table. When the ground station receives the remote sensing information required by the user equipment, it queries from the satellite remote sensing information table. If it can be queried, the satellite with the required remote sensing information is the source node, and the satellite above the user equipment is the destination node. According to the satellite node and link information, the routing table from the source node to the destination node is calculated through the routing algorithm and sent to the satellite. After receiving the routing table, the satellite transmits remote sensing data.

[0013] Furthermore, the specific steps of step S3 are as follows:

[0014] Step S31: Collect historical data and perform data preprocessing;

[0015] Step S32: using an LSTM neural network prediction algorithm based on EMD and attention mechanism to predict traffic;

[0016] Step S33: Use mean absolute error and root mean square error to perform prediction evaluation.

[0017] Furthermore, the specific process of step S32 is as follows:

[0018] Original traffic decomposition: Let the original sequence be X(t), and the original sequence is decomposed into a finite number of IMFs cj and a remainder term r n ; The original sequence expression is:

[0019] ;

[0020] in, c j ( t ) represents the intrinsic mode function, r n ( t ) represents a remainder;

[0021] The calculation method of the original series standard deviation is:

[0022] ;

[0023] in, h j,i ( t ) is the intrinsic mode function of iteration i, h j,i-1 ( t ) is the intrinsic mode function iterated i-1 times, j represents the order of the function, i represents the number of iterations, and N represents the total number of iterations;

[0024] Short correlation component prediction: The LSTM node structure includes an input gate, a forget gate, and an output gate; the forget gate is used to control the retention of the state information of the previous neuron node, and the input gate is used to control the entry into the node state C t The calculation formula is:

[0025] ;

[0026] ;

[0027] Among them, x t Represents the input of data, h t-1 is the output of the hidden layer at the previous moment, b f is the forget gate bias term, W f is the forget gate weight, W i is the input gate weight, b i is the input gate bias term, σ is the Sigmoid activation function, and the output after processing is f t , determines the degree of discarding historical information;

[0028] The output gate is used to update the unit state and determine the output. The calculation method of the new state information of the node is:

[0029] ;

[0030] ;

[0031] Among them, b c is the output gate bias term; W c is the output gate weight, C t-1 is the cell state at the previous moment, i t is the input gate value;

[0032] The output state h of the hidden layer t and the output value o t It is the final calculation step of the node. The output value and output status are expressed as follows:

[0033] ;

[0034] ;

[0035] Among them, W o is the output layer weight, h t-1 is the hidden state of the previous time step, b o is the output gate bias term.

[0036] The final mapping relationship is as follows:

[0037] ;

[0038] ;

[0039] Among them, W y and b y are the parameters of LSTM to be trained, representing weights and offsets, y n Represents x after LSTM mapping n Output:

[0040] Pay attention to the degree of hidden state at time T and calculate the output sequence at time t:

[0041] ;

[0042] Among them, W h is the hidden layer weight, b h is the hidden layer bias term; represents the hidden state of the previous time sequence, represents the hidden state at time t; the proportion of attention at time t to the total attention at all times is:

[0043] ;

[0044] Finally, all components are added together as the output of the entire model.

[0045] Furthermore, in step S4, the input of the improved ant colony algorithm is the satellite network topology and the service request, and the output is the path between the source node and the destination node that satisfies the delay-bandwidth constraint; the improved ant colony algorithm specifically includes the following steps:

[0046] Step S41: determining the satellite network topology in the time slice to which the current moment belongs;

[0047] Step S42: Determine the source node s and destination node d of the service request, and initialize parameters, including the number of ants m, the number of iterations N=0, and the maximum number of iterations N maxInitialize pheromone concentration, pheromone volatility coefficient, pheromone factor, inspiration factor and taboo table;

[0048] Step S43: Set the source node as the ant's current node and add it to the taboo table, N=N+1;

[0049] Step S44: constructing a solution space, the ant calculates the forwarding probability of the optional link of the current node according to the state transition rule and selects the next hop, and adds the selected node to the taboo table;

[0050] Step S45: the ant determines whether the current node is the destination node; if the current node is the destination node, the path finding is successful; if the current node is not the destination node, the path finding continues;

[0051] Step S46: The ant determines whether the allowed set of the current node is an empty set; if the allowed set is an empty set, the path finding fails; otherwise, execute step S44;

[0052] Step S47: After the ant reaches the destination node d, it returns along the original route to update the pheromone;

[0053] Step S48: Determine whether the termination condition is met; if N=N max , then find the optimal path according to the objective function, otherwise set the taboo table = Φ and go to step S43.

[0054] Further, in step S44, represents the probability of an ant moving from node i to an optional node j, which is calculated as:

[0055] ;

[0056] in, represents the set of next nodes that ant k can choose; s represents the neighboring nodes of node i at time t; represents the pheromone concentration of link i, j at time t; represents the heuristic degree of link i, j at time t; represents the pheromone concentration of link i, s at time t; represents the inspiration degree of link i, s at time t; α and β are the pheromone factor and inspiration factor respectively.

[0057] Furthermore, the improved ant colony algorithm is optimized in satellite routing, and the specific process is as follows:

[0058] Set the initial value of link pheromone:

[0059] ;

[0060] Where H is a constant, is the remaining available bandwidth of link i, j;

[0061] The following optimizations are made to the heuristic function:

[0062] ;

[0063] in, is the cost function constructed in the routing model;

[0064] The exploration probability q0 is introduced. When the ant selects the next hop, it chooses the path according to the link selection probability calculated by the original formula or makes a random selection with an average probability.

[0065] Introduction γ As the link congestion index, it is equal to the ratio of the link's used bandwidth to the remaining bandwidth. γ 0 is the determined link congestion threshold, and the attribute expression is:

[0066] ;

[0067] in, is the link used bandwidth, is the link bandwidth, is the remaining bandwidth of the link.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] The method provided by the present invention provides two remote sensing information collection methods on the one hand, which can meet the communication needs of different user devices, and on the other hand can effectively predict business traffic, pre-allocate resources, reduce transmission delay, and improve the overall performance of the communication system. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a schematic diagram of the system model in the present invention.

[0071] Figure 2 It is a flow chart of the present invention.

[0072] Figure 3 The flowchart of the improved ant colony algorithm in the present invention. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0074] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0075] Embodiment 1, as Figure 1 and Figure 2 As shown, a method for optimizing satellite network transmission based on service prediction for integrated communication and remote sensing is provided in one embodiment of the present invention, comprising the following steps:

[0076] Step S1: construct a system model including satellites, user equipment and ground stations;

[0077] In the step S1, a communication and remote sensing integrated network transmission system model is constructed. The remote sensing task usually involves multiple collaborative operation platforms, and its complete operation process includes: the ground station issues operation instructions to the communication and remote sensing integrated satellite system according to the measurement and control instructions; the communication and remote sensing integrated satellite system is responsible for the collection of remote sensing data; the system then performs preliminary processing on the satellite data to generate pre-processed remote sensing data; such pre-processed remote sensing data is transmitted to the ground station via the satellite-to-ground link; the ground station further transfers the pre-processed remote sensing data to the data computing center, which processes and generates the final remote sensing product and forwards it to the remote sensing user equipment.

[0078] The core idea of ​​this system architecture is to use the remote sensing satellite network as a remote sensing data perception node, and seamlessly access the communication satellite network with all-weather, global coverage without blind spots. The communication satellite network serves as the high-speed relay "information trunk" of the remote sensing satellite network, realizing the rapid integration and distribution of multi-source remote sensing data, thereby breaking through the limitations of the traditional remote sensing satellite system's long revisit time and limited regional coverage. At the same time, intersatellite links can further accelerate the transmission and processing of remote sensing data and shorten the remote sensing response cycle. Compared with traditional remote sensing satellite systems, this system does not need to increase the cost of repeated system construction and maintenance, but can also improve the efficiency of information application and enhance the timeliness of remote sensing.

[0079] Step S2: Two methods of collecting remote sensing data by integrated communication and remote sensing satellites are proposed, and the method of collecting remote sensing data can be selected according to user equipment requirements;

[0080] In step S2, the two communication and remote sensing integrated satellites collect remote sensing data in the following manner:

[0081] The first method: the user device uploads the location to be photographed. The ground station uses the satellite ephemeris information and the star selection algorithm to determine the satellite closest to the area (shooting star) and the next closest backup star (backup shooting star). The shooting information is uploaded to enable the shooting star to perform the operation. When the shooting star performs attitude maneuvers for shooting, it uploads the shooting task to the ground station and other functions are suspended. After the shooting star finishes the operation, it sends information to the ground station to indicate that it has finished the shooting task, and other functions return to normal. After that, the ground station uses the satellite routing algorithm, with the shooting star as the source node and the satellite above the user device as the destination node. According to the satellite node and link information, the routing algorithm is used to calculate the routing table from the source node to the destination node and transmit it to the satellite. After receiving the routing table, the satellite transmits remote sensing data.

[0082] The second type: the user device uploads the location of the specific time at which the shooting is required, and the user device transmits the time information and space information of the required remote sensing information to the ground station. The satellite uses flat scanning to collect remote sensing data. Each satellite assigns time attributes and space attributes to the collected remote sensing information (similar to the remote sensing information above Changbai Mountain at 8:20 on December 13, 2023). After that, each satellite transmits the time attributes and space attributes of its own remote sensing information block to the ground station to form a remote sensing information table. When the ground station receives the remote sensing information required by the user device, it queries from the satellite remote sensing information table. If it can be queried, the satellite with the required remote sensing information is called a resource satellite (source node), and the satellite above the user device is called a co-transmission satellite (destination node). According to the satellite node and link information, the routing table from the source node to the destination node is calculated through the routing algorithm and transmitted to the satellite. After receiving the routing table, the satellite transmits remote sensing data.

[0083] Step S3: predicting network traffic data within a future time interval through historical remote sensing data; pre-allocating resources through advance prediction of services, reducing real-time resource allocation delay, and improving the overall performance of the communication system;

[0084] The specific steps of step S3 are as follows:

[0085] Collect historical data, perform data preprocessing, use the LSTM neural network prediction algorithm based on EMD and attention mechanism to predict traffic, and use the mean absolute error (MAE) and root mean square error (RMSE) to evaluate the prediction.

[0086] In order to make the prediction model adapt to the self-similarity and long-range correlation of remote sensing business data, the EMD decomposition method is used to decompose the original traffic sequence into multiple single-frequency and independent subsequences, and then each subsequence is predicted through the LSTM neural network model. On this basis, the attention mechanism is introduced to assign different weights to different hidden layer states through mapping weights and learning parameter matrices to strengthen the influence of important information. Finally, the output of the attention mechanism is added as the prediction output. The specific process is as follows:

[0087] Original traffic decomposition: Let the original sequence be X(t), and the original sequence is decomposed into a finite number of IMFs cj and a remainder term r n ; The original sequence expression is:

[0088] ;

[0089] in, c j ( t ) represents the intrinsic mode function, r n ( t ) represents a remainder;

[0090] The calculation method of the original series standard deviation is:

[0091] ;

[0092] in, h j,i ( t ) is the intrinsic mode function of iteration i, h j,i-1 ( t ) is the intrinsic mode function iterated i-1 times, j represents the order of the function, i represents the number of iterations, and N represents the total number of iterations;

[0093] Short correlation component prediction: The LSTM node structure includes an input gate, a forget gate, and an output gate, which respectively realize the input of information, the update of historical status, and the output; the forget gate is used to control the retention of the state information of the previous neuron node, and the input gate is used to control the entry into the node state C t The calculation formula is:

[0094] ;

[0095] ;

[0096] Among them, x t Represents the input of data, h t-1 is the output of the hidden layer at the previous moment, b f is the forget gate bias term, Wf is the forget gate weight, W i is the input gate weight, b i is the input gate bias term, σ is the Sigmoid activation function, and the output after processing is f t , determines the degree of discarding historical information;

[0097] The output gate is used to update the unit state and determine the output. The calculation method of the new state information of the node is:

[0098] ;

[0099] ;

[0100] Among them, b c is the output gate bias term; W c is the output gate weight, C t-1 is the cell state at the previous moment, i t is the input gate value;

[0101] The output state h of the hidden layer t and the output value o t It is the final calculation step of the node. The characteristic of LSTM is to adjust h t To reduce the loss of overall information, thereby alleviating the network's forgetfulness. The output value and output state are expressed as follows:

[0102] ;

[0103] ;

[0104] Among them, W o is the output layer weight, h t-1 is the hidden state of the previous time step, b o is the output gate bias term.

[0105] The final mapping relationship is as follows:

[0106] ;

[0107] ;

[0108] Among them, W y and b y are the parameters of LSTM to be trained, representing weights and offsets, y n Represents x after LSTM mapping n Output:

[0109] Information weight evaluation: When the input time series is output through LSTM, the changing state of the traffic at different times, the hidden state and memory value of the output can be obtained, and the output value after being affected by the past time characteristics can be obtained. Since the information at different times has different weights on the state prediction at the current time, it is necessary to evaluate the importance of information generated at different times. The attention mechanism can focus on more concerned information among a lot of input information and reduce the attention of other information, thereby improving the efficiency and accuracy of the prediction.

[0110] Calculating the output sequence at time t requires paying attention to the degree of the hidden state at time T:

[0111] ;

[0112] Among them, W h is the hidden layer weight, b h is the hidden layer bias term; represents the hidden state of the previous time sequence, represents the hidden state at time t; the proportion of attention at time t to the total attention at all times is:

[0113] ;

[0114] After the attention mechanism, each component is added as the output of the entire model. By calculating the proportion of attention at different times, the trained model can not only grasp the temporal characteristics of the traffic sequence, but also strengthen the weight of the moment that has a greater impact on the output value through the attention mechanism, thereby finely depicting the temporal characteristics of network traffic and improving the prediction accuracy of the model.

[0115] The evaluation indicators of time series forecasting: mean absolute error (MAE) and root mean square error (RMSE) are:

[0116] ;

[0117] ;

[0118] in, represents the predicted value, is the actual data, and n is the number of samples.

[0119] In step S3, the steps of the satellite routing algorithm are as follows:

[0120] The satellite network operation cycle is divided into several time slices at equal intervals. A corresponding network topology state graph is constructed for each time slice, and an undirected full graph G(V, L, P) is used to describe the topology state graph. V represents the set of connected nodes in the satellite network, L represents the set of connected links between nodes, and P represents the set of paths from the source node to the destination node d. i and j are nodes in the network, L(i, j) represents the communication link between node i and node j, and P(s, d) represents a routing path from the source node to the destination node. Each communication link has a corresponding link transmission delay delay(i, j). The routing algorithm is to find the routing path P(s, d) from the source node s to the destination node d, and minimize the link transmission delay and satisfy certain constraints.

[0121] Minimize the delay objective function:

[0122] ;

[0123] Constraints:

[0124] ;

[0125] Among them, D is the maximum delay required for the path from the source node to the destination node to transmit a data packet, and B is the minimum bandwidth required for the link to transmit the data packet.

[0126] Step S4: Using the improved ant colony algorithm to calculate the routing table based on the predicted service traffic to balance the link load.

[0127] In step S4, the input of the improved ant colony algorithm is the satellite network topology and the service request, and the output is the path between the source node and the destination node that meets the delay and bandwidth constraints; the improved ant colony algorithm specifically includes the following steps (see Figure 3 ):

[0128] Step S41: determining the satellite network topology in the time slice to which the current moment belongs;

[0129] Step S42: Determine the source node s and destination node d of the service request, and initialize parameters, including the number of ants m, the number of iterations N=0, and the maximum number of iterations N max Initialize pheromone concentration, pheromone volatility coefficient, pheromone factor, inspiration factor and taboo table;

[0130] Step S43: Set the source node as the ant's current node and add it to the taboo table, N=N+1;

[0131] Step S44: constructing a solution space, the ant calculates the forwarding probability of the optional link of the current node according to the state transition rule and selects the next hop, and adds the selected node to the taboo table;

[0132] Step S45: The ant determines whether the current node j is the destination node d; if the current node is the destination node, the path finding is successful; if the current node is not the destination node, the path finding continues;

[0133] Step S46: The ant determines whether the allowed set of the current node is an empty set; if the allowed set is an empty set, the path finding fails; otherwise, execute step S44;

[0134] Step S47: After the ant reaches the destination node d, it returns along the original route to update the pheromone;

[0135] Step S48: Determine whether the termination condition is met; if N=N max , then find the optimal path according to the objective function, otherwise set the taboo table = Φ and go to step S43.

[0136] The basic ant colony algorithm achieves optimization through the adaptation and collaboration process of ants, with the following assumptions:

[0137] (1) When foraging, ants sense pheromones along the path and leave new pheromones along the path they have walked.

[0138] (2) The ant’s internal model determines its response to the environment;

[0139] (3) Individual ants are random and independent, but the group is self-organized and orderly. The adaptation process is a positive feedback process. The shorter the path, the greater the probability that the path will be chosen by the ants, and the greater the concentration of accumulated pheromones, which will attract more ants to choose this path. The collaboration process is the acquisition and exchange of information between ants, that is, exchanging information between candidate solutions to optimize performance.

[0140] Assume that there are m ants in the ant colony, and initially all edges have the same pheromone concentration, which is a constant C. At time t, for a single ant k, represents the probability of an ant moving from node i to an optional node j, which is calculated as:

[0141] ;

[0142] in, represents the set of next nodes that ant k can choose; s represents the neighboring nodes of node i at time t; represents the pheromone concentration of link i, j at time t; represents the heuristic degree of link i, j at time t. The heuristic function can generally be written as the inverse of the link distance; represents the pheromone concentration of link i, s at time t; It represents the inspiration degree of link i, s at time t; α and β are pheromone factor and inspiration factor respectively. In the process of ant pathfinding, the importance of pheromone concentration and inspiration function is adjusted by setting the values ​​of α and β.

[0143] Link selection probability From the calculation formula, we can see that the more concentrated the pheromone is and the higher the inspiration degree is, the greater the probability of being selected. All nodes passed by the ants will be added to the taboo table to avoid loops and affect the convergence of the algorithm.

[0144] When the ant completes a cycle, it needs to update the pheromone on the link between node i and node j. The link pheromone at time t + 1 is:

[0145] ;

[0146] ;

[0147] in, represents the link pheromone at time t, It represents the increment of pheromone on the link from node i to node j in this cycle of m ants; It indicates the amount of pheromone left by the kth ant on the link from i to j in this cycle; Indicates the volatility coefficient of pheromone, with a value between 10-1;

[0148] There are three models of ant pheromone release, corresponding to pheromone update There are also three ways to calculate , namely the ant-period model, the ant-quantity model, and the ant-density model. The ant-quantity model and the ant-density model use local network information, while the ant-period model uses global network information. Taking the ant-period model as an example, the pheromone increment is related to the path of this search, which is a global information update. Its calculation is as follows:

[0149] ;

[0150] Where Q represents the pheromone intensity (constant); L K Represents the length of the path ant k passes through in this cycle (the sum of the lengths of all links passed in the path).

[0151] The improved ant colony algorithm is used to optimize satellite routing. The specific process is as follows:

[0152] Pheromone initialization: In the basic ant colony algorithm, the pheromone concentration of all links is set equal at the beginning. When directly applied to the satellite network, the traffic situation in the link is not considered, which is not conducive to load balancing, and the algorithm converges to the optimal path slowly. In order to take into account network congestion and reduce the number of algorithm iterations, the initial value of the link pheromone is set according to the bandwidth usage in the network:

[0153] ;

[0154] Where H is a constant, is the remaining available bandwidth of link i, j; the initial pheromone concentration of the link is proportional to the remaining available bandwidth, which attracts ants to links with more available resources and reduces the pheromone accumulation on paths carrying more services.

[0155] Heuristic function: The heuristic function in the basic ant colony algorithm only considers the distance between nodes but not the load information. If it is directly applied to the satellite network, a large number of services will be concentrated on the shorter paths. Therefore, the following optimizations are made:

[0156] ;

[0157] in, is the cost function constructed in the routing model; the smaller the cost value, the larger the value of the heuristic function, the higher the corresponding link selection probability, and the more the ants tend to choose the path with low cost.

[0158] State transition rule: During the execution of the ant colony algorithm, after repeated searches, ants will select the optimal path with a probability very close to 1, while the probability of selecting other paths is greatly reduced, and the solution space cannot be further searched. In order to overcome this stagnant behavior, q0 is introduced. When the ants select the next hop, they either select the path according to the link selection probability calculated by the original formula, or randomly select with an average probability. Here, q0 is the exploration probability. By adjusting the size of q0, two purposes can be achieved. First, the optimal path is still selected with a probability (1-q0) that is larger but no longer close to 1. Second, when the performance of the original optimal path deteriorates and a new optimal path appears, the ants will go to other paths with an exploration probability of q0, and then they can quickly find and gather on the new optimal path to avoid stagnation in the search.

[0159] Pheromone update rule: In the basic ant colony algorithm, pheromone consists of two parts, namely the residue of the previous cycle and the increment of this cycle. The expression of the increment of pheromone on the link from node i to node j in one cycle is For the already congested paths that the ants have walked, the pheromone will still increase, making the ants less sensitive to load changes and converging more slowly. γAs the link congestion index, it is equal to the ratio of the link's used bandwidth to the remaining bandwidth. γ 0 is the determined link congestion threshold, and the attribute expression is:

[0160] ;

[0161] in, is the link used bandwidth, is the link bandwidth, is the remaining bandwidth of the link.

[0162] when γ When it is small, the pheromone update is calculated according to the basic ant colony algorithm. γ When it is larger, , that is, reduce the iterative increase of pheromones on the path that has been congested, and reduce the probability of selecting this path.

[0163] In the embodiment of the present invention, this embodiment can meet the communication needs of different user devices on the one hand, and on the other hand, it can also effectively predict business traffic, pre-allocate resources, reduce transmission delay, and improve the overall performance of the communication system.

[0164] Embodiment 2, this embodiment provides a satellite communication routing strategy based on user equipment demand prediction, which can predict the demand of user equipment in different areas for satellite communication, such as voice calls, video calls, data transmission, etc. According to the user equipment demand prediction, the satellite communication link is dynamically adjusted to connect the area with large user equipment demand with the satellite communication link to improve network performance and user equipment experience. The specific implementation process can refer to the implementation process of the method steps in Embodiment 1, which will not be repeated here.

[0165] Embodiment 3, this embodiment provides a satellite communication routing strategy based on emergency communication needs, which can predict the emergency communication needs of different regions, such as emergency communication needs when natural disasters such as earthquakes, floods, and typhoons occur. According to the emergency communication needs, the satellite responsible for communication in the disaster area is connected to the corresponding communication link to ensure smooth information communication between the disaster area and the outside world. The specific implementation process can refer to the implementation process of the method steps in Embodiment 1, which will not be repeated here.

[0166] The above are only preferred embodiments of the present invention. It should be pointed out that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These should also be regarded as the protection scope of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A method for optimizing satellite network transmission based on business forecasting, characterized in that: The following steps are involved: Step S1: construct a system model including satellites, user equipment and ground stations; Step S2: Propose two methods of collecting remote sensing data by integrated satellite communication and remote sensing, and select one method of collecting remote sensing data; Step S3: predicting network traffic data within a future time interval through historical remote sensing data; Step S4: Calculate the routing table using an improved ant colony algorithm based on the predicted service traffic; In step S2, the two communication and remote sensing integrated satellites collect remote sensing data in the following manner: The first method: the user device uploads the location to be photographed, and the ground station uses the star selection algorithm to determine the shooting star and the backup shooting star based on the satellite ephemeris information, and uploads the shooting information to enable the shooting star to perform the operation. When the shooting star performs attitude maneuvers for shooting, it uploads the execution of the shooting task to the ground station; after the shooting star finishes the operation, it sends information to the ground station, and then the ground station uses the satellite routing algorithm, with the shooting star as the source node and the satellite above the user device as the destination node. According to the satellite node and link information, the routing algorithm is used to calculate the routing table from the source node to the destination node and transmit it to the satellite. After receiving the routing table, the satellite transmits remote sensing data; The second type: the user equipment transmits the time information and space information of the required remote sensing information to the ground station. The satellite uses flat scanning to collect remote sensing data. Each satellite assigns time attributes and space attributes to the collected remote sensing information. After that, each satellite transmits the time attributes and space attributes of its own remote sensing information block back to the ground station to form a remote sensing information table. When the ground station receives the remote sensing information required by the user equipment, it queries from the satellite remote sensing information table. If it can be queried, the satellite with the required remote sensing information is the source node, and the satellite above the user equipment is the destination node. According to the satellite node and link information, the routing table from the source node to the destination node is calculated by the routing algorithm and sent to the satellite. After receiving the routing table, the satellite transmits the remote sensing data; The specific steps of step S3 are as follows: Step S31: Collect historical data and perform data preprocessing; Step S32: using an LSTM neural network prediction algorithm based on EMD and attention mechanism to predict traffic; Step S33: using mean absolute error and root mean square error to perform prediction evaluation; The specific process of step S32 is as follows: Original traffic decomposition: Let the original sequence be X(t), and the original sequence is decomposed into a finite number of IMFs cj and a remainder term r n ; The original sequence expression is: ； in, c j ( t ) represents the intrinsic mode function, r n ( t ) represents a remainder; The calculation method of the original series standard deviation is: ； in, h j,i ( t ) is the intrinsic mode function of iteration i, h j,i-1 ( t ) is the intrinsic mode function iterated i-1 times, j represents the order of the function, i represents the number of iterations, and N represents the total number of iterations; Short correlation component prediction: The LSTM node structure includes an input gate, a forget gate, and an output gate; the forget gate is used to control the retention of the state information of the previous neuron node, and the input gate is used to control the entry into the node state C t The calculation formula is: ； ； Among them, x t Represents the input of data, h t-1 is the output of the hidden layer at the previous moment, b f is the forget gate bias term, W f is the forget gate weight, W i is the input gate weight, b i is the input gate bias term, σ is the Sigmoid activation function, and the output after processing is f t , determines the degree of discarding historical information; The output gate is used to update the unit state and determine the output. The calculation method of the new state information of the node is: ； ； Among them, b c is the output gate bias term; W c is the output gate weight, C t-1 is the cell state at the previous moment, i t is the input gate value; The output state h of the hidden layer t and the output value o t It is the final calculation step of the node. The output value and output status are expressed as follows: ； ； Among them, W o is the output layer weight, h t-1 is the hidden state of the previous time step, b o is the output gate bias term; The final mapping relationship is as follows: ； ； Among them, W y and b y are the parameters of LSTM to be trained, representing weights and offsets, y n Represents x after LSTM mapping n Output: Pay attention to the degree of hidden state at time T and calculate the output sequence at time t: ； Among them, W h is the hidden layer weight, b h is the hidden layer bias term; represents the hidden state of the previous time sequence, represents the hidden state at time t; the proportion of attention at time t to the total attention at all times is: ； Finally, all components are added together as the output of the entire model; In step S4, the input of the improved ant colony algorithm is the satellite network topology and the service request, and the output is the path between the source node and the destination node that meets the delay bandwidth constraint; the improved ant colony algorithm specifically includes the following steps: Step S41: determining the satellite network topology in the time slice to which the current moment belongs; Step S42: Determine the source node s and destination node d of the service request, and initialize parameters, including the number of ants m, the number of iterations N=0, and the maximum number of iterations N max Initialize pheromone concentration, pheromone volatility coefficient, pheromone factor, inspiration factor and taboo table; Step S43: Set the source node as the ant's current node and add it to the taboo table, N=N+1; Step S44: constructing a solution space, the ant calculates the forwarding probability of the optional link of the current node according to the state transition rule and selects the next hop, and adds the selected node to the taboo table; Step S45: the ant determines whether the current node is the destination node; if the current node is the destination node, the path finding is successful; if the current node is not the destination node, the path finding continues; Step S46: The ant determines whether the allowed set of the current node is an empty set; if the allowed set is an empty set, the path finding fails; otherwise, execute step S44; Step S47: After the ant reaches the destination node d, it returns along the original route to update the pheromone; Step S48: Determine whether the termination condition is met; if N=N max , then find the optimal path according to the objective function, otherwise set the taboo table = Φ and go to step S43; In step S44, represents the probability of an ant moving from node i to an optional node j, which is calculated as: ； in, represents the set of next nodes that ant k can choose; s represents the neighboring nodes of node i at time t; represents the pheromone concentration of link i, j at time t; represents the heuristic degree of link i, j at time t; represents the pheromone concentration of link i, s at time t; represents the inspiration degree of link i, s at time t; α and β are the pheromone factor and inspiration factor respectively; The improved ant colony algorithm is used to optimize satellite routing. The specific process is as follows: Set the initial value of link pheromone: ； Where H is a constant, is the remaining available bandwidth of link i, j; The following optimizations are made to the heuristic function: ； in, is the cost function constructed in the routing model; By introducing the exploration probability q0, when the ant selects the next hop, it chooses the path according to the link selection probability calculated by the original formula or makes a random selection with an average probability; Introduction γ As the link congestion index, it is equal to the ratio of the link's used bandwidth to the remaining bandwidth. γ 0 is the determined link congestion threshold, and the attribute expression is: ； in, B used is the link used bandwidth, B base is the link bandwidth, B residual is the remaining bandwidth of the link.

Images