A computing power scheduling method for multi-satellite cooperative application

Abstract

The application discloses a kind of computing power scheduling methods of multi-star coordination application: step 1, ground station generates undirected connection diagram according to satellite constellation topological structure and resource state;Step 2, ground station receives user instruction, and carries out task computing power demand estimation;Step 3, ground station determines execution area on undirected connection diagram using sliding window mode, and executes task;Step 4, the center node of execution area is calculated, and it is used as the source satellite of execution information collection task;Step 5, source satellite executes information collection task;Step 6, using double factor sorting scheduling strategy, task data distribution from source satellite to other satellite nodes in area is completed;Step 7, satellite node returns result to source satellite after completing task processing;Step 8, source satellite sends result to ground station after collecting it.The application provides a kind of strategy for multi-star cooperative processing and resource dynamic scheduling under the condition of single-satellite limited capacity, reduces the processing delay of on-board data.

Description

Technical Field

[0001] This invention belongs to the field of satellite communication technology, specifically relating to a computing power scheduling method for multi-satellite collaborative applications. Background Technology

[0002] The performance improvements, payload diversification, and large-scale constellations of satellites will inevitably lead to an exponential increase in on-board data, making it difficult for existing operating methods to meet the demands for massive, low-latency on-orbit information transmission. On the other hand, the dramatic increase in the number of satellites has also created enormous computing power resources for numerous onboard processors. If onboard computing resources can be integrated, and key information can be extracted and processed onboard through multi-satellite collaboration, it would be possible to reduce the amount of data transmitted downlink, trading computation for bandwidth, while simultaneously improving service timeliness. This is also an effective means of solving high latency and downlink bottlenecks.

[0003] However, traditional methods often use a single satellite overhead for information collection and task processing. With the deployment of mega-constellations and the application of emerging technologies such as computing networks, the single-satellite model will inevitably shift to a multi-satellite collaborative model. The selection of which satellites to use for collaboration, the number of satellites to use, and how to distribute task data between satellites will become crucial issues that need to be addressed and resolved in the multi-satellite collaborative model. Summary of the Invention

[0004] The purpose of this invention is to propose a computing power scheduling method for multi-satellite collaborative applications, so as to solve the problems of limited single-satellite resources and untimely service response in the existing technology.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A computing power scheduling method for multi-satellite collaborative applications specifically includes the following steps: Step 1: The ground station generates an undirected connectivity graph based on the satellite constellation topology and resource status; Step 2: The ground station receives user instructions and estimates the task's computing power requirements; Step 3: The ground station selects the over-the-top satellite region on the undirected connectivity graph and uses a sliding window method to estimate the computing power requirements to finally determine the execution area of ​​the task. Step 4: Calculate the central node of the execution area and use it as the source satellite for the information acquisition task; Step 5: The source satellite performs information acquisition tasks; Step 6: Employ a two-factor sorting and scheduling strategy to distribute task data from the source satellite to other satellite nodes in the region; Step 7: Process the mission data completed by the satellites within the region and return the processed results to the source satellite; Step 8: The source satellite collects the results and sends them to the ground station.

[0006] Furthermore, step 1 includes the following sub-steps: Step 11: Satellite Node S i Obtain information about the current status of the resources being processed. P i Inter-satellite link status information L i And inform the ground station. Among them, P i Characterized by the number of bits that can be processed per second, the unit of measurement is... bps , L i Characterized by the number of bits that can be transmitted per second, the unit of measurement is... bps ; Step 12: Ground station calculates satellite nodes S i computing power C i .in, C i = min ( P i , L i ), L i = ( L i,top + L i,down + L i,left + L i,right ) / 4, L i,top , L i,down , L i,left , L i,right Representing satellite nodes respectively S i The transmission rates of two interconnected inter-satellite links on the same orbit and two interconnected inter-satellite links on different orbits. Step 13: The ground station generates an undirected connectivity graph. G Among them, the nodes in the undirected connected graph G i Represents its corresponding satellite node S i In an undirected connected graph, the connectivity between nodes is represented by the connectivity between satellite nodes, and the numerical values ​​of nodes in the undirected connected graph are represented by the values ​​of satellite nodes. S i computing power C i .

[0007] Furthermore, the method for estimating the task's computing power requirement in step 2 is as follows: The ground station compares the data type and algorithm complexity of the current task with those of related tasks in the historical database to find the most similar task. Then, based on the computing power requirement of the most similar task and the data volume of the current task and the most similar task, it estimates the computing power requirement of the current task. CT .

[0008] Furthermore, step 3 includes the following sub-steps: Step 3 includes the following sub-steps: Step 31: The ground station selects the over-the-head satellite region Ω1 on the undirected connectivity graph, where the size of Ω1 is... W 1a × W 1b ,in, W 1a , W 1b These represent the number of satellite nodes in the horizontal and vertical directions, respectively. Step 32: Read the task latency requirements from the instruction. T deadline This allows us to estimate the minimum computing power required for the task. C min = CT / T deadline Initialize the size of the sliding window to W 2a × W 2b ,in, W 2a =2、 W 2b =2. Step 33: Set variables num =0; Step 34: Within the continuous region Ω1, calculate the total computing resources within the selected window using a sliding window approach. C sum And determine whether it satisfies... ,in, α For adjustment value, satisfying α >1, a typical value can be 2. If satisfied, mark the window region as Ω2 and jump to step 4. If not satisfied, if num =1, then jump to step 36, if num If the value is 2, then proceed to step 37; otherwise, execute step 35. Step 35: W 2a =W 2a +1, num = num +1. Check if the condition is met. W 2a ≤ W 1a If the condition is met, proceed to step 34; otherwise, proceed to step 36. Step 36: W 2a = W 2a -1、 W 2b = W 2b +1, num = num +1. Check if the condition is met. W 2b ≤ W 1b If the condition is met, proceed to step 34; otherwise, proceed to step 37. Step 37: W 2a = W 2a +1, check if the condition is met. W 2a ≤ W 1a and W 2b ≤ W 1b If the condition is met, proceed to step 33; otherwise, proceed to step 38. Step 38: Discard the task and proceed to step 2.

[0009] Furthermore, step 4 includes the following sub-steps: Step 41: Establish a coordinate system. Select any satellite node within this region as the origin of the two-dimensional coordinate system. For each additional hop in the direction of the out-of-orbit right satellite, x Increment the axis coordinate by 1, otherwise subtract 1; for each jump in the direction of the previous star in the same orbit... y Increment the axis coordinate by 1, otherwise subtract 1; Step 42: Calculate the centroid of this region. The calculation method is as follows: x c = , y c = .in( x c , y c ( ) represents the centroid coordinates. x i , yi Satellite nodes S i exist x axis, y Coordinates on the axis C i For satellite nodes S i The computing power value, n This represents the number of satellite nodes within the region. Step 43: Select the satellite node closest to the centroid as the regional center node; Furthermore, step 6 includes the following sub-steps: Step 61: Satellite nodes within region Ω2 S i The shortest inter-satellite link hop count between the source satellite node and the source satellite node is denoted as hop i ; Step 62: Calculate the satellite nodes within region Ω2 S i sort value R i .in, , β , γ These are the weight values ​​for the reciprocal of the hop count and the computing power value, respectively. Typical values ​​can be taken as follows: β =0.5、 γ =0.5; Step 63: According to R i The task data is distributed from the source satellite node to other satellite nodes in descending order of size.

[0010] The beneficial effects of this invention compared to the prior art are: (1) In view of the current characteristics of limited computing power resources of single satellite and strong correlation between service response time and computing power, this invention designs a satellite computing power scheduling method for multi-satellite collaborative application, realizes dynamic allocation of satellite network resources and multi-satellite collaborative processing, improves the resource utilization rate of satellite network and reduces service response latency; (2) Traditional methods often use a single satellite overhead for information collection and task processing. With the deployment of mega-constellations and the application of emerging technologies such as computing networks, the current single-satellite mode will inevitably shift to a multi-satellite collaborative mode in the future. In response to this trend, this invention provides a specific solution to problems such as satellite area selection, information collection satellite selection, and inter-satellite task data distribution, providing technical support for the future application of computing networks in large-scale space-based constellations; In summary, the method of this invention provides a solution for satellite computing power scheduling in multi-satellite collaborative applications through steps such as satellite region selection, information acquisition satellite selection, and inter-satellite mission data distribution. This aims to improve the resource utilization and overall efficiency of satellite networks and reduce the processing latency of on-board data. Attached Figure Description

[0011] Figure 1 This is a flowchart of the present invention; Figure 2 It is an undirected connected graph; Figure 3 A diagram illustrating region selection and window sliding; Figure 4 A schematic diagram of the established coordinate system; Detailed Implementation The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings.

[0012] like Figure 1 As shown, this invention takes the LEO constellation as an example to provide a computing power scheduling method for multi-star collaborative applications. The implementation steps are as follows: Step 1: The ground station generates an undirected connectivity graph based on the LEO topology and resource status; Step 1 includes the following sub-steps: Step 11, LEO node S i Obtain information about the current status of the resources being processed. P i Inter-satellite link status information L i And inform the ground station. Among them, P i Characterized by the number of bits that can be processed per second, the unit of measurement is... bps , L i Characterized by the number of bits that can be transmitted per second, the unit of measurement is... bps ; Step 12: Ground station calculates satellite nodes S i computing power C i .in, C i = min ( P i , L i ), L i = ( L i,top + L i,down +L i,left + L i,right ) / 4, L i,top , L i,down , L i,left , L i,right Representing satellite nodes respectively S i The transmission rates of two interconnected inter-satellite links on the same orbit and two interconnected inter-satellite links on different orbits. Step 13: The ground station generates an undirected connectivity graph. G Among them, the nodes in the undirected connected graph G i Represents its corresponding satellite node S i In an undirected connected graph, the connectivity between nodes is represented by the connectivity between satellite nodes, and the numerical values ​​of nodes in the undirected connected graph are represented by the values ​​of satellite nodes. S i computing power C i .

[0013] In this embodiment, the satellite S 1. The current remaining processing capacity is 500Mbps, and the transmission capacities of the four inter-satellite links are respectively... L 1,top =5Gbps L 1,down =3Gbps L 1,left =3Gbps L 1,right =3Gbps. Then P 1 = 500Mbps L 1 = 4.5Gbps. C 1= min ( P 1, L 1) = 500Mbps. And so on, calculating satellite nodes. S 2. S 3. S 4. ... computing power value C 2. C 3. C 4. ... and generate, based on the connectivity between satellites, such as... Figure 2 The wireless connection diagram is shown.

[0014] Step 2: The ground station receives user instructions and estimates the task's computing power requirements. Specifically, firstly, the ground station establishes a computing power demand sample database through extensive simulations and field measurements. This database records the computing power requirements of commonly used tasks under different task types, data types, data volumes, and algorithm complexities. Then, the ground station constructs a three-dimensional scalar based on the task type, data type, and algorithm complexity of the current task. It then uses Euclidean distance calculations within the computing power demand sample database to find the task item that best matches the current task. Finally, based on the size of the computing power demand in the database, the size of the current task's data volume, and the number of items in the database, the computing power demand of the current task is estimated. CT .

[0015] In this embodiment, the current task is to detect ships in a certain sea area with a swath width of 1km × 1km, using the YOLO algorithm. The computational power requirement estimation method for the current task is as follows: First, the ground station uses parameters such as task type (object detection), data type (optical image), and algorithm complexity (complexity of the YOLO algorithm) to find the task with the smallest Euclidean distance between the current task and the sample database, and marks it as... Task i Then, the computing power requirement of the current task is estimated by using the ratio of the two data volumes. CT =Data volume of the current task / Task i Data volume × Task i The computing power requirements.

[0016] Step 3: The ground station selects the over-the-top satellite region on the undirected connectivity graph and uses a sliding window method to estimate the computing power requirements to finally determine the execution area of ​​the task. Step 3 includes the following sub-steps: Step 31: The ground station selects the over-the-head satellite region Ω1 on the undirected connectivity graph, where the size of Ω1 is... W 1a × W 1b , W 1a , W 1b These represent the number of satellites in the horizontal and vertical directions, respectively. Figure 3 As shown in this implementation case W 1a =5、 W 1b =4; Step 32: Read the task latency requirements from the instruction. T deadline This allows us to estimate the minimum computing power required for the task. C min = CT / T deadline And initialize the size of the sliding window to... W 2a × W 2b ,in, W 2a =2、 W 2b =2; Step 33: Set variables num =0; Step 34: Within the continuous region Ω1, calculate the total computing resources within the selected window using a sliding window approach. C sum And determine whether it satisfies... ,in, α For adjustment value, satisfying α >1, a typical value can be 2. If satisfied, mark the window region as Ω2 and jump to step 4. If not satisfied, if num =1, then jump to step 36, if num If the value is 2, then proceed to step 37; otherwise, execute step 35. Step 35: W 2a = W 2a +1, num = num +1. Check if the condition is met. W 2a ≤ W 1a If the condition is met, proceed to step 34; otherwise, proceed to step 36. Step 36: W 2a = W 2a -1、 W 2b = W 2b +1, num = num +1. Check if the condition is met. W 2b ≤ W 1b If the condition is met, proceed to step 34; otherwise, proceed to step 37. Step 37: W 2a = W 2a +1, check if the condition is met. W 2a ≤ W 1a and W2b ≤ W 1b If the condition is met, proceed to step 33; otherwise, proceed to step 38. Step 38: Discard the task and proceed to step 2.

[0017] In this implementation case, it is assumed that the execution region that ultimately meets the conditions is as follows: Figure 3 The region Ω2 shown W 2a =3、 W 2b =3.

[0018] Step 4: Calculate the central node of the execution area and use it as the source satellite for the information acquisition task; Step 4 includes the following sub-steps: Step 41: Establish as follows Figure 4 The coordinate system shown is used in this implementation example. S 14 The origin is at coordinates (0,0). S 12 , S 13 , S 17 , S 18 , S 19 , S 22 , S 23 , S 24 The coordinates are (0,2), (0,1), (1,2), (1,1), (1,0), (2,2), (2,1), (2,0); Step 42: Calculate the centroid of the region. In this implementation case, it is assumed that... C 12 , C 13 , C 14 , C 17 , C 18 , C 19 , C 22 , C 23 , C 24 The values ​​are 0.5, 1, 0.8, 0.6, 0.5, 0.7, 0.2, 0.8, and 0.4, respectively, in Gbps. Therefore: x c = = ≈0.84, y c = = ≈0.93. The centroid coordinates of this region are (0.84, 0.93); Step 43: Select the satellite node closest to the centroid as the regional center node, i.e., select the satellite. S 18 It serves as the central node of this region; Step 5: Satellite execution S 18 Information collection task; Step 6: Employ a two-factor sorting and scheduling strategy to complete the self-satellite scheduling. S 18 Distribute mission data to other satellite nodes in the region; Step 6 includes the following sub-steps: Step 61: Satellite nodes within the measurement area S 12 , S 13 , S 14 , S 17 , S 19 , S 22 , S 23 , S 24 With information acquisition satellite nodes S 18 The shortest inter-satellite link hop counts are: 2, 1, 2, 1, 1, 2, 1, 2; Step 62: Calculate satellite nodes within the region S i sort value R i .in, β , γ The value is 0.5. C i , hop i After normalizing the reciprocals, we get , , , , , , , , , The values ​​are: 0.1, 0.2, 0.16, 0.12, 0.14, 0.04, 0.16, and 0.08, respectively. , , , , , , , They are respectively: , , , , , , , Thus obtain R 12 =0.0917、 R 13 =0.1833、 R 14 =0.1217、 R 17 =0.1433、 R 19 =0.1533、 R 22 =0.0617、 R 23 =0.1633、 R 24 =0.0817; Step 63: According to R i Complete the process from source satellite to satellite node in descending order. S i Inter-satellite mission data distribution. That is, satellites. S 18 The priority for distributing data to other satellite nodes is S 13 , S 23 , S 19 , S 17 , S 14 , S 12 , S 24 , S 22 ; Step 7: Process the satellite data within the region to complete the mission, and return the processed results to the source satellite; In this implementation case, it is assumed that the satellite node S 18 Selected S 13 , S 23 , S 19 , S 17 , S 14 If the five nodes work together, then S 13 , S 23 , S 19 , S 17 , S 14 After completing the ship inspection based on remote sensing image data, the results were sent to the source satellite. S 18 Since the results are from shipboard detection, it can be assumed that each satellite node sends data to the source satellite. S 18 The amount of data is negligible.

[0019] Step 8: The source satellite collects the results and sends them to the ground station.

[0020] Compared to the traditional single-satellite processing mode, the multi-satellite collaborative mode proposed in this invention effectively reduces task processing latency. Taking the acquisition of 100 Gbit of data from a single satellite as an example, the differences between the method of this invention and traditional methods are illustrated.

[0021] In traditional single-satellite processing mode, if on-board processing is used, the processing latency mainly depends on the satellite's computing power. As explained above, satellite nodes... S 12 , S 13 , S 17 , S 18 , S 19 , S 22 , S 23 , S 24 computing power C 12 , C 13 , C 14 , C 17 ,C 18 , C 19 , C 22 , C 23 , C 24 The values ​​are 0.5, 1, 0.8, 0.6, 0.5, 0.7, 0.2, 0.8, and 0.4, respectively, in Gbps. Then, select a satellite node. S 13 The processing latency is the lowest at that time. T delay =100 / 1=100s; Select satellite node S 22 The processing latency is the lowest at that time. T delay =100 / 0.2 =500s. Therefore, when processing a single star, T delay The value is located in the interval [100s, 500s].

[0022] In the multi-satellite collaborative mode proposed in this invention, assuming an inter-satellite transmission rate of 2Gbps, then... S 18 , S 13 , S 23 , S 19 , S 17 , S 14 The latency for a single satellite to process this 100Gbps (total latency = processing latency + transmission latency) are: 200, 150, 175, 193, 217, and 225 seconds, respectively. S 18 The task data is distributed based on the reciprocal ratio of the latency of each satellite. S 18 , S 13 , S 23 , S 19 , S 17 , S 14 The assigned task data are: 15.8, 21, 18.1, 16.4, 14.6, 14.1, in gigabit (Gbit). Therefore, the task processing latency is equal to... S 18 , S 13 ,S 23 , S 19 , S 17 , S 14 The maximum value of the delay of these nodes, i.e. T delay= 31.6s. Compared to traditional single-satellite processing, this significantly reduces latency.

[0023] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. A computing power scheduling method for multi-satellite collaborative applications, characterized in that, Specifically, the steps include the following: Step 1: The ground station generates an undirected connectivity graph based on the satellite constellation topology and resource status; Step 2: The ground station receives user instructions and estimates the task's computing power requirements. The ground station compares the data type and algorithm complexity of the current task with the data type and algorithm complexity of related tasks in the historical database to find the most similar task item. Then, based on the computing power requirement of the most similar task item and the data volume of the current task and the most similar task item, it estimates the computing power requirement of the current task. ; Step 3: The ground station selects the over-the-head satellite region on the undirected connectivity graph and uses a sliding window approach to determine the execution region of the task based on the estimated computing power requirements; this includes the following sub-steps: Step 31: The ground station selects the over-the-head satellite region on the undirected connectivity map. , The size is ,in, These represent the number of satellite nodes in the horizontal and vertical directions, respectively. Step 32: Read the task latency requirements from the instruction. This allows us to estimate the minimum computing power required for the task. Initialize the size of the sliding window to ,in, , ; Step 33: Set variables ; Step 34: In the overhead satellite region Within the selected window, a sliding window method is used to calculate the total computing resources within that window. And determine whether it satisfies If the conditions are met, then mark the region as . And jump to step 4; if not satisfied, if Then proceed to step 36, if If the condition is met, proceed to step 37; otherwise, execute step 35. For adjustment value, satisfying ; Step 35: Determine if the condition is met. If the condition is met, proceed to step 34; otherwise, proceed to step 36. Step 36: Determine if the condition is met. If the condition is met, proceed to step 34; otherwise, proceed to step 37. Step 37: Determine whether the condition is met. and If the condition is met, proceed to step 33; otherwise, proceed to step 38. Step 38: Discard the task and proceed to step 2; Step 4: Calculate the central node of the execution area and use it as the source satellite for the information acquisition task; Step 5: The source satellite performs information acquisition tasks; Step 6: Employ a two-factor sorting and scheduling strategy to distribute task data from the source satellite to other satellite nodes in the region; the two-factor sorting and scheduling strategy is as follows: Step 61: Area Satellite nodes within The shortest inter-satellite link hop count between the source satellite node and the source satellite node is denoted as ; Step 62: Calculate the region Inner satellite node sort value ;in, , , These are the weight values ​​for the reciprocal of the hop count and the computing power value, respectively; where... For satellite nodes The computing power value; Step 63: According to The task data is distributed from the source satellite node to other satellite nodes in descending order of size. Step 7: Process the mission data completed by the satellites within the region and return the processed results to the source satellite; Step 8: The source satellite collects the results and sends them to the ground station.

2. The computing power scheduling method for multi-satellite collaborative applications as described in claim 1, characterized in that, Step 1 includes the following sub-steps: Step 11: Satellite Node Obtain information about the current status of the resources being processed. Inter-satellite link status information And inform the ground station; among them, Characterized by the number of bits that can be processed per second, the unit of measurement is... , Characterized by the number of bits that can be transmitted per second, the unit of measurement is... ; Step 12: Ground station calculates satellite nodes computing power ;in, , , Representing satellite nodes respectively The transmission rates of two interconnected inter-satellite links on the same orbit and two interconnected inter-satellite links on different orbits. Step 13: The ground station generates an undirected connectivity graph. ; where nodes in an undirected connected graph Represents its corresponding satellite node In an undirected connected graph, the connectivity between nodes is represented by the connectivity between satellite nodes, and the numerical values ​​of nodes in the undirected connected graph are represented by the values ​​of satellite nodes. computing power .

3. The computing power scheduling method for multi-satellite collaborative applications as described in claim 1, characterized in that, In step 34, the adjustment value .

4. The computing power scheduling method for multi-satellite collaborative applications as described in claim 1, characterized in that, The method for selecting information acquisition satellite nodes in step 4: Step 41: Establish a coordinate system; Specifically, this involves selecting any satellite node within the region as the origin of the two-dimensional coordinate system, and for each additional hop in the direction of the out-of-orbit right satellite, ... Increment the axis coordinate by 1, otherwise subtract 1; for each jump in the direction of the previous star in the same orbit, Increment the axis coordinate by 1, otherwise subtract 1; Step 42: Calculate the centroid of this region. The calculation method is as follows: = , = ;in Using the centroid coordinates, Satellite nodes exist x axis, Coordinates on the axis For satellite nodes The computing power value, This represents the number of satellite nodes within the region. Step 43: Select the satellite node closest to the centroid as the center node of the region and use it as the source satellite node for information acquisition.

Images