Task-oriented multi-layer low-orbit satellite giant satellite base clustering networking method

Abstract

The invention relates to a task-oriented multi-layer low-orbit satellite giant satellite base clustering networking method, which comprises the following steps of: 1, initializing satellite constellation data which comprises load resources of satellites, geographic positions and orbit position information of the satellites and inter-satellite link establishment conditions of the satellites, and after a task is generated, performing clustering networking on the satellite constellation data; receiving the task data by a control center located on the ground or a high-orbit satellite; step 2, the control center establishes a task model according to satellite resources required by the task, and establishes a satellite model according to orbit parameters of the satellite and satellite load capacity data; establishing a specific task and satellite model establishment mechanism; and 3, determining limiting conditions according to different task types, and the like. The method has the advantages that mathematical modeling is carried out on the tasks and the satellites, the cooperation relation between the satellites and the matching relation between the satellites and task requirements are combined through the node enhancement graph, and a foundation is laid for task-oriented on-demand clustering.

Description

Technical Field

[0001] The invention belongs to the field of mobile communication technology, and in particular relates to a task-oriented multi-layer low-orbit satellite giant constellation clustering networking method. Background Art

[0002] Satellite communication networks have wide coverage and high flexibility, and have become an important part of the sixth generation of mobile communication technology (6G). With the construction of satellite constellations, many satellite constellation construction plans such as Starlink and OneWeb have been launched one after another, forming a multi-layer giant satellite constellation with ubiquitous satellites in space. With the maturity of inter-satellite link technology and the enhancement of satellite payload capacity, satellite on-board self-organizing cluster networking has become a key research direction. In the future, the current situation of independently operating satellite constellations with isolated resources, functions and communication systems will gradually be broken. Flexible networking of distributed constellations based on on-board self-organization will become a trend. Its core concept is that heterogeneous small satellites improve network performance through inter-satellite collaboration to form a large "virtual satellite". When mission requirements change or face emergencies, satellites already deployed in orbit can quickly form a network to form new capabilities without launching new satellites. This approach will significantly improve the response speed and flexibility to sudden demands.

[0003] The multi-layer low-orbit satellite giant constellation network refers to a layered satellite network composed of a large number of low earth orbit (LEO) satellites. The satellites in the network are usually in orbits of various altitudes and inclinations, and have inter-satellite links. The networking forms between satellites are also different, including satellite constellation networking and star cluster companion flying networking. In addition, there are both military and civilian satellites in the multi-layer low-orbit satellite giant constellation network. The communication systems and on-board payloads between these satellites are also different. Some satellites may have cross-domain communication capabilities, such as carrying cross-domain routing protocols and multi-beam laser links, which can communicate across domains with satellites with different communication systems and protocols. Although the multi-layer low-orbit satellite giant constellation network has great application prospects, it also faces many challenges. Due to its huge scale, highly complex inter-satellite collaboration, complex management between heterogeneous satellites, and dynamically changing network topology, how to efficiently manage the entire multi-layer low-orbit satellite giant constellation network has become a difficult problem to solve.

[0004] Clustering networking is a popular strategy for achieving flexible network management in ground networks and an effective way to further improve the performance of satellite networks. It is usually divided into two types: clustering networking technology for wireless sensor networks (WSNs) and clustering networking technology for mobile ad hoc networks (MANETs). In WSNs, the main purpose of network clustering technology is to extend the network life under limited energy conditions. The representative article of this type of clustering technology is the LEACH clustering algorithm proposed in "Energy-efficient communication protocol for wireless microsensor networks". This algorithm randomly selects cluster heads and members and balances energy consumption by rotating cluster heads. In MANET, the mobility between nodes is the main consideration of network clustering technology. Most of this type of research is based on weight-based clustering algorithms. The selection of weights is very flexible and can be based on mobility, QoS, connectivity, etc. Typical MANET networking strategies are also widely used in vehicle-mounted ad hoc networks (VANETs) and unmanned aerial vehicle ad hoc networks (UAVs). The article "QMM-VANET: An efficient clustering algorithmbased on QoS and monitoring of malicious vehiclesin vehicular ad hoc networks" proposes the QMM-VANET clustering algorithm, which achieves low-latency and high-stability network clustering in VANET by comprehensively considering QoS requirements, distrust value parameters and mobility constraints. Chinese invention patent application number CN201910197312.9 discloses an invention named a cluster drone self-organizing network clustering method, the method comprising: initializing a cluster head, the cluster head broadcasting cluster head invitation information to surrounding nodes; according to the region-based division method of bird flocks, the cluster head divides the surrounding area, and divides the cluster structure in the self-organizing network into: cluster head area, node attraction area and node exclusion area; wherein, the nodes in the cluster head area are cluster head candidate nodes; the node attraction area includes: a gateway area and a member area, the nodes in the gateway area are gateway nodes in the cluster, and the nodes in the member area are member nodes in the cluster; the nodes in the node exclusion area include: idle nodes and nodes in other clusters; when an idle node receives invitation information broadcast by multiple cluster heads within a preset first time period, it marks its own status according to the signal strength of the received invitation information to complete the node deployment. The method uses the region-based division method of bird flocks to mark the node status according to the signal strength of the received invitation information to provide a more stable cluster structure.

[0005] There is currently little research on clustering networking algorithms in satellite networks. In the past, the academic community has tended to study hierarchical satellite network structures: "MLSR: A novel routing algorithm for multilayered satellite IPnetworks" proposed a satellite clustering architecture of MLSN, which consists of a MEO / GEO satellite and all LEO satellites within its coverage area, effectively reducing the complexity of on-board routing table calculation and communication load. Chinese invention patent application number CN201811249213.2 discloses an invention entitled "Communication method, device and system based on satellite network, the method comprising: a user terminal receives a detection signal sent by a management satellite on a management channel, the management satellite manages one or more service satellites; the user terminal sends a breathing signal to the management satellite, the breathing signal carries information of the user terminal, and the information of the user terminal is used to determine the service satellite information serving the user terminal. Due to the operations such as the user terminal receiving the detection signal and sending the breathing signal, the management satellite or the ground station can uniformly dispatch the service satellite to serve the user terminal. When the service satellite is switched, the step of negotiation between satellites can be omitted, signaling overhead is saved, and the satellite communication experience of the user terminal is smooth. That is, in this method, a management satellite and multiple service satellites constitute a satellite "super cell". This concept of dividing the satellite network into multiple "super cells" is consistent with the cluster networking technology. The concept of cluster networking is also mentioned in satellite networks based on SDN, such as the article "Dynamic SDN controllerplacement in a LEO constellation satellite Chinese invention patent application number CN202110812489.2 discloses an invention named "An SDN-based integrated ground-ground adaptive dynamic QoS routing method", which includes the following steps: establishing a hierarchical clustered network model based on SDN; establishing network resource mapping; establishing a multi-constrained QOS adaptive routing algorithm SDN-AD. The invention can effectively reduce control overhead and improve transmission efficiency by reducing long-distance transmission and shortest-distance clustering of control packets, and formulate multi-constrained QoS problems as optimization problems with the minimum transmission cost as the goal. It can effectively calculate the transmission costs of links of different categories, shield the differences between satellites and ground networks at different levels, so as to better adapt to network changes, thereby providing services that meet different service quality requirements. The present invention solves the optimization problem and realizes adaptive routing.In addition, the article “Reliable and Low-Overhead Clustering in LEO Small Satellite Networks” proposed a distributed online solution based on alliance game theory, which achieved highly reliable and low-overhead satellite clustering networking and was verified in a single-layer satellite network, fully proving that the clustering networking idea has great application prospects in satellite networks.

[0006] Different from the cluster structure of ground cluster networking, the cluster network structure of the future multi-layer low-orbit satellite giant constellation network is usually a cross-layer and cross-domain three-dimensional cluster structure because the satellites in the network are usually in orbits of different altitudes and inclinations, and the networking forms between satellites are diverse. There are cross-layer intersatellite links between satellite layers. These characteristics make the cluster network structure of the multi-layer low-orbit satellite giant constellation network usually a cross-layer and cross-domain three-dimensional cluster structure.

[0007] The ground clustering networking algorithm is based on a homogeneous network structure, and nodes are often equal. In wireless ad hoc networking scenarios, mobile ad hoc networks, and Internet of Vehicles scenarios, the cluster structures formed by clustering networking technology are all flat clusters. Because the ground clustering networking algorithm relies on the network topology, it cannot meet the needs of the multi-layer low-orbit satellite giant constellation network to establish a three-dimensional cluster structure across layers.

[0008] UAV self-organizing networks and multi-layer low-orbit satellite giant constellation networks have similarities in three-dimensional network clusters, but the clustering of UAV self-organizing networks mostly uses the signal strength between UAVs as the basis for clustering. Since satellites in satellite networks use high-power transmitting equipment and often use directional antennas, the control of signal loss during transmission is more stable than that of UAV self-organizing networks, making the signal strength factor have less impact on satellite networks. Therefore, clustering networking based on signal strength cannot be directly used in satellite networks. UAVs in UAV self-organizing networks can change their positions according to networking requirements, but it is difficult for satellites to do this. In addition, the scale of UAV networks is much smaller than that of satellite giant constellation networks. Its clustering algorithm is targeted at small-scale networks and is usually relatively simplified and difficult to adapt to the scalability requirements of large-scale satellite clustering networks.

[0009] In summary, the current clustering methods in satellite networks are mainly used for networking of homogeneous and homogeneous satellites, aiming to optimize network management. The clustering methods in existing technologies ignore the collaborative potential between different satellites and different constellations, and cannot flexibly and efficiently utilize the increasingly abundant resource payloads on board to form functional clusters across constellations and resource domains. Summary of the invention

[0010] The purpose of the present invention is to overcome the problems existing in the above-mentioned prior art and to provide a task-oriented multi-layer low-orbit satellite giant constellation clustering networking method.

[0011] The task-oriented multi-layer low-orbit satellite giant constellation cluster networking method comprises the following steps:

[0012] Step 1. Initialize satellite constellation data, including satellite payload resources, satellite geographic location and orbital information, and satellite inter-satellite link establishment status. When a mission is generated, the mission data is received by the control center located on the ground or in a high-orbit satellite.

[0013] Step 2. The control center establishes a mission model based on the on-board resources required by the mission, and a satellite model based on the satellite's orbital parameters and on-board payload capacity data; specific mission and satellite model establishment mechanism;

[0014] Step 3. Determine constraints based on different task types;

[0015] Step 4. Model the cooperation between satellites and the matching relationship between satellites and tasks as a Markov transition probability model and establish a transition probability matrix;

[0016] Step 5. Obtain candidate satellite members and candidate cluster head satellites through the transfer probability matrix obtained in step 4, and calculate the functional cluster matching the current task based on the cluster head satellite as the center and the constraints in step 3;

[0017] Step 6. Upload the networking data to the satellite network. After receiving the networking data, the satellite network establishes a functional cluster through the inter-satellite link status notification mechanism and saves the intra-cluster routing table.

[0018] Furthermore, in step 1, the ground or high-orbit satellite control center initializes and maintains various data of the satellite constellation through ephemeris information and satellite information, including the satellite's payload resources, the satellite's geographic location and orbital information, and the satellite's inter-satellite link establishment status. When a task is generated, the task requirement will first be transmitted to the ground or high-orbit satellite control center, and the control center will execute subsequent steps to divide the satellite task cluster.

[0019] Furthermore, in step 2, the control center establishes a mission model based on the on-board resources required by the mission, and establishes a satellite model based on the satellite's orbital parameters and on-board payload capacity data; the specific mission and satellite model establishment mechanism is as follows:

[0020] Step 2.1. Task model establishment mechanism

[0021] The ad hoc task is defined as a set TASK, which is composed of the following formula (1):

[0022] TASK = {task 1 ,task 2 ,…,task TN}......Formula (1),

[0023] Among them: task i represents a subtask in an ad hoc task, TN represents the number of subtasks, and the definition of subtask tk is as follows (2)-(3):

[0024]

[0025] Where: type i Indicates the type of subtask, time i represents the execution time required for the subtask, φ i Represents the set of load resources required by the subtask, where: represents the value of the j-th resource requirement of the i-th subtask, RN represents the total number of resource requirements, The larger the value of , the greater the demand of the subtask for this type of resource. If the subtask has no demand for this type of resource, then φ i1 The value of is 0;

[0026] Step 2.2 Satellite model establishment mechanism

[0027] A satellite is defined as the set s i , which is composed of the following formulas (4)-(5):

[0028]

[0029]

[0030] L i ={a,e,θ,Ω,g,v}......(6),

[0031] N i ={s 1 ,s 2 ,…,s N ,N=|N i |}......(7),

[0032] Where: id i Indicates the satellite number. Indicates the cluster head satellite number of the cluster to which the satellite belongs, tw i represents the time window of the satellite covering the mission area, δ i Represents the satellite's onboard payload resource set, represents the value of the j-th type of payload resources on the i-th satellite, and One-to-one correspondence, L i Indicates the six orbital numbers of the satellite, a represents the semi-major axis, e represents the eccentricity, θ represents the orbital inclination, Ω represents the longitude of the ascending node, g represents the perigee angular distance, ν represents the true anomaly angle of the perigee epoch, and N irepresents the set of neighbor satellites of satellite i, and N represents the number of neighbor satellites.

[0033] Furthermore, in step 3, the restriction conditions are determined according to different task types, specifically:

[0034] For the functional satellite cluster that meets the mission requirements, the satellite functional cluster overhead formula is proposed as the satellite cluster constraint, and a typical constraint method based on cluster size, resource redundancy rate, total number of hops and mission execution time is given. The constraint conditions are shown in the following equations (8)-(12):

[0035] S(CL)=α·K+β·R(CL)+γ·H(CL)......(8),

[0036] CL={CH,CM 1 ,CM 2 ,…,CM K-1},K= |CL| ......(9),

[0037]

[0038]

[0039] Where: S(CL) represents the cluster management cost, CL represents the set of satellite function cluster members, CH represents the cluster head satellite, CM represents the cluster member satellite, K represents the size of the satellite cluster, R(CL) represents the redundancy of resources within the cluster, H(CL) represents the total number of routing hops from the member satellite to the cluster head satellite, α, β, γ are weight parameters used to adjust the impact of various factors on S(CL); Equations (11) and (12) are the resource and time constraints of the satellite function cluster. For a satellite function cluster that meets the mission requirements, the sum of the resources required by all cluster member satellites for a certain mission is greater than the total demand for the resources by the mission, and the time window of any member satellite meets the satellite coverage time requirements of the subtask assigned to the satellite.

[0040] Furthermore, in step 4, the cooperation between satellites and the matching relationship between satellites and tasks are modeled as a Markov transition probability model, and a transition probability matrix is ​​established, which specifically includes the following steps:

[0041] Step 4.1 Establish a Markov transition probability model

[0042] The process of satellites being classified into different satellite clusters is regarded as the transfer of the cooperation state between the satellite and different satellites, which is modeled as a Markov process. Let S t represents the set of all possible cooperative states in the satellite network, state represents the specific cooperative state of a node at time t, and iTransition to state j The probability model is as shown in equations (13)-(16):

[0043] P ij (t) = Pr(S t+1 =state j |S t =state i )......(13),

[0044]

[0045]

[0046] Among them: the transition probability satisfies the normalization, boundedness and symmetry conditions, and the value of the transition probability depends on the possibility of cooperation between satellites due to mission requirements. Specifically, a weighted graph based on the satellite topology structure is constructed to characterize the cooperation relationship between satellites and the matching relationship between satellites and mission requirements.

[0047] In this step, G(V,E) is used to represent the satellite structure topology, where V represents the satellite set and E represents the satellite inter-satellite link (ISL) set. Based on the graph G, a node enhancement graph is proposed: Among them, V st represents the satellite node, V at Represents the subtask node in TK, connecting the edges of structure points It is called a structural edge, which represents the possible ISL set between satellites within the communication range, and the edge connecting the structural point and the attribute point It is called the attribute edge, which represents the matching relationship between the satellite and the subtask. The weights of the structure edge and the attribute edge are ω i,j It indicates that a weight definition is given based on the mission required resources and on-board resources as follows (17):

[0048]

[0049] For structural edges, ω i,j Indicates the similarity of resources between two structural nodes and is positively correlated with the number of common resources between the structural nodes. For attribute edges, when the number of matching resources between the structural node and the attribute node is not 0, ω i,j The value of is positive, ω baseSt and ω baseAt is the correction value of the weight of the structural edge and the attribute edge. The size of the correction value is set according to actual needs. baseSt and ω baseAt The values ​​of are 10 and 100 respectively. In addition, the attribute nodes are not connected to each other;

[0050] In the node-enhanced graph, if multi-hop neighbors Matched with the same subtask, share a new neighbor, the attribute node They become two-hop neighbors, shortening their effective distance in the graph and increasing the possibility of them being classified into the same cluster. The transition probability of structural edges and attribute edges is defined as follows:

[0051]

[0052] Among them, N 1 Represents the number of neighbor structure points of the node, N 2 The number of neighbor attribute points representing the node, by combining equation (17) with equation (18), we get the transition probability matrix between all structure points and attribute points (19):

[0053]

[0054] Step 4.2 Solve the Markov clustering algorithm (MCL) based on the transition probability matrix P

[0055] The MCL algorithm simulates the random walk process through matrix multiplication. The MCL algorithm captures the relationship between related nodes in the clustering process and obtains clusters from the graph structure by expanding and contracting the transition probability matrix P. The MCL algorithm is as follows:

[0056] Step 4.2.1 The transfer probability matrix P is expanded, that is, P = P ε , simulate the multi-step random walk process to spread the transition probability. Among them, ε is the expansion parameter, which is greater than 1 and represents the number of random walk steps. In order to add inertia to the transition probability and ensure that the transition probability matrix P can finally converge, a self-loop is added to each structural point and attribute node, and the weight of the self-loop is fixed to ω baseAt ;

[0057] Step 4.2.2 performs an expansion operation to further increase the transition probability and suppress the low transition probability. The expansion operation includes performing an r-th power operation on each element of the transition probability matrix P and then normalizing the obtained elements:

[0058]

[0059] Where r>1 represents the expansion parameter, which is used to control the granularity of clustering. The larger the expansion parameter, the more cluster structures will be obtained.

[0060] Step 4.2.3 Repeat steps 4.2.1 and 4.2.2 until the matrix converges and obtains a stable probability distribution. In the converged transfer probability matrix, the high-value elements in each column will be concentrated in the same row as the high-value elements in other columns, indicating that the elements in this row belong to the same cluster.

[0061] Further, in step 5, the candidate satellite members and candidate cluster head satellites are obtained through the probability transfer matrix obtained in step 4, and the functional cluster matching the current task is calculated based on the cluster head satellite as the center and the restriction conditions in step 3, which specifically includes the following steps:

[0062] Step 5.1 Add the task TK′ of selecting cluster head satellite nodes according to the ephemeris information, and construct the node enhancement graph with all satellite nodes as structural points and task node TK′ as the only attribute point. and the probability transformation matrix P'. and P′ are input into the MCL algorithm to obtain the candidate set CH of potential cluster head satellites cd ;

[0063] Step 5.2: Take candidate set CH cd Each cluster head CH i As the center, reduce the maximum hop range hop max The edge weights ω of all satellites other than i,j , and all edge weights ω of the satellites whose subtask requirements have been met i,j ;

[0064] Step 5.3 constructs a new probability transition matrix P, which includes all satellites and all subtasks (except TK′) nodes. Input this matrix into the MCL algorithm to obtain a functional cluster candidate CL i Repeat steps 3 and 4 to obtain a candidate set of functional clusters CL;

[0065] Step 5.4: For each functional cluster candidate CL i Calculate the cluster management cost S(CL) and return the function cluster candidate with the minimum S(CL) as the final function cluster.

[0066] Compared with the prior art, the present invention has the following superior technical effects:

[0067] 1. The existing clustering networking algorithm only considers the similarity between nodes due to the topological structure, and cannot cluster on demand for tasks. The task-oriented multi-layer low-orbit satellite giant star clustering networking method described in the present invention combines the collaborative relationship between satellites and the matching relationship between satellites and task requirements through a node enhancement graph by mathematically modeling tasks and satellites, thereby laying a foundation for task-oriented on-demand clustering. It not only reduces the pressure of multi-layer giant star network management through clustering networking, but also makes more efficient and flexible use of the rich on-board payload resources of the satellite network.

[0068] 2. The task-oriented multi-layer low-orbit satellite giant star clustering networking method described in the present invention models the clustering networking problem as a Markov transition probability process. The state transition in the Markov process only depends on the current probability distribution, which provides greater flexibility for network clustering, so that the satellite clustering network can dynamically adjust the clustering strategy according to the current mission requirements, satellite status and environmental number; while the traditional clustering algorithm based on fixed rules or predictions lacks flexible response to real-time environmental changes and mission requirements. In addition, the clustering method based on the Markov transition probability matrix can effectively evaluate the collaboration potential between satellites based on a global perspective, thereby making it easier to achieve global optimal clustering than traditional methods.

[0069] 3. The task-oriented multi-layer low-orbit satellite giant star clustering networking method described in the present invention is not limited to homogeneous and same-layer satellite networks when constructing node enhancement graphs and transition probability matrices, but considers the potential collaboration objects of each satellite, thereby achieving support for heterogeneous and cross-layer satellite networks.

[0070] 4. The task-oriented multi-layer low-orbit satellite giant constellation clustering networking method described in the present invention can effectively realize the task-oriented clustering networking of satellite networks, reduce cluster size, resource occupancy rate, total network hops and network construction time, and improve the on-board resource utilization rate of the multi-layer low-orbit satellite giant constellation; actual verification results show the adaptability of the method described in the present invention to the multi-layer low-orbit satellite giant constellation. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 This is a schematic diagram of the mission-oriented multi-layer low-orbit satellite giant constellation cluster networking scenario;

[0072] Figure 2 The flowchart of the mission-oriented multi-layer low-orbit satellite giant constellation cluster networking method is provided;

[0073] Figure 3 Schematic diagram of the task-oriented satellite function cluster generation algorithm flow chart;

[0074] Figures 4a-4d Schematic diagram of the impact of the number of different mission resource requirements on satellite clusters. DETAILED DESCRIPTION

[0075] In order to more clearly understand the above-mentioned objectives, features and advantages of the present invention, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0076] Example

[0077] like Figure 1 -4, the task-oriented multi-layer low-orbit satellite giant constellation cluster networking method of the present invention comprises the following steps:

[0078] Step 1. Initialize satellite constellation data, including satellite payload resources, satellite geographic location and orbital information, and satellite inter-satellite link establishment status. When a mission is generated, the mission data is received by the control center located on the ground or in a high-orbit satellite;

[0079] Step 2. The control center establishes a mission model based on the on-board resources required by the mission, and a satellite model based on the satellite's orbital parameters and on-board payload capacity data; specific mission and satellite model establishment mechanism;

[0080] Step 3. Determine constraints based on different task types;

[0081] Step 4. Model the cooperation between satellites and the matching relationship between satellites and tasks as a Markov transition probability model and establish a transition probability matrix;

[0082] Step 5. Obtain candidate satellite members and candidate cluster head satellites through the transfer probability matrix obtained in step 4, and calculate the functional cluster matching the current task based on the cluster head satellite as the center and the constraints in step 3;

[0083] Step 6. Upload the networking data to the satellite network. After receiving the networking data, the satellite network establishes a functional cluster through the inter-satellite link status notification mechanism and saves the intra-cluster routing table.

[0084] As a specific embodiment of the present invention, in step 1, the ground or high-orbit satellite control center initializes and maintains various data of the satellite constellation through ephemeris information and satellite information, including satellite payload resources, satellite geographic location and orbital position information, and satellite inter-satellite link establishment status. When a task is generated, the task requirement will first be transmitted to the ground or high-orbit satellite control center, and the control center will perform subsequent steps to divide the satellite task cluster.

[0085] As a specific embodiment of the present invention, in step 2, the control center establishes a task model according to the on-board resources required by the task, and establishes a satellite model according to the orbital parameters of the satellite and the on-board payload capacity data; the specific task and satellite model establishment mechanism is as follows:

[0086] Step 2.1. Task model establishment mechanism

[0087] The ad hoc task is defined as a set TASK, which is composed of the following formula (1):

[0088] TASK = {task 1 ,task 2 ,…,task TN}......(1),

[0089] Among them: task i represents a subtask in an ad hoc task, TN represents the number of subtasks, and the definition of subtask tk is as follows (2)-(3):

[0090] task i ={type i ,φ i ,time i}......(2),

[0091]

[0092] Where: type i Indicates the type of subtask, time i represents the execution time required for the subtask, φ i Represents the set of load resources required by the subtask, where: represents the value of the j-th resource requirement of the i-th subtask, RN represents the total number of resource requirements, The larger the value of ≥0, the greater the demand of the subtask for this type of resource. If the subtask has no demand for this type of resource, then The value of is 0;

[0093] Step 2.2 Satellite model establishment mechanism

[0094] A satellite is defined as the set s i , which is composed of the following formulas (4)-(5):

[0095]

[0096]

[0097] L i ={a,e,θ,Ω,g,v}......(6),

[0098] N i ={s 1 ,s 2 ,…,s N ,N=|N i |}......(7),

[0099] Where: id i Indicates the satellite number. Indicates the cluster head satellite number of the cluster to which the satellite belongs, tw i represents the time window of the satellite covering the mission area, δ i Represents the satellite's onboard payload resource set, represents the value of the j-th type of payload resources on the i-th satellite, and One-to-one correspondence, L i Indicates the six orbital numbers of the satellite, a represents the semi-major axis, e represents the eccentricity, θ represents the orbital inclination, Ω represents the longitude of the ascending node, g represents the perigee angular distance, ν represents the true anomaly angle of the perigee epoch, and N i represents the set of neighbor satellites of satellite i, and N represents the number of neighbor satellites.

[0100] As a specific embodiment of the present invention, in step 3, the restriction conditions are determined according to different task types, specifically:

[0101] For the functional satellite cluster that meets the mission requirements, the satellite functional cluster overhead formula is proposed as the satellite cluster constraint, and a typical constraint method based on cluster size, resource redundancy rate, total number of hops and mission execution time is given. The constraint conditions are shown in the following equations (8)-(12):

[0102] S(CL)=α·K+β·R(CL)+γ·H(CL)......(8),

[0103] CL={CH,CM 1 ,CM 2 ,…,CM K-1},K= |CL| ......(9),

[0104]

[0105]

[0106] Where: S(CL) represents the cluster management cost, CL represents the set of satellite function cluster members, CH represents the cluster head satellite, CM represents the cluster member satellite, K represents the size of the satellite cluster, R(CL) represents the redundancy of resources within the cluster, H(CL) represents the total number of routing hops from the member satellite to the cluster head satellite, α, β, γ are weight parameters used to adjust the impact of various factors on S(CL); Equations (11) and (12) are the resource and time constraints of the satellite function cluster. For a satellite function cluster that meets the mission requirements, the sum of the resources required by all cluster member satellites for a certain mission is greater than the total demand for the resources by the mission, and the time window of any member satellite meets the satellite coverage time requirements of the subtask assigned to the satellite.

[0107] As a specific embodiment of the present invention, in step 4, the cooperation between satellites and the matching relationship between satellites and tasks are modeled as a Markov transition probability model, and a transition probability matrix is ​​established, which specifically includes the following steps:

[0108] Step 4.1 Establish a Markov transition probability model

[0109] The process of satellites being classified into different satellite clusters is regarded as the transfer of the cooperation state between the satellite and different satellites, which is modeled as a Markov process. Let S t represents the set of all possible cooperative states in the satellite network, state represents the specific cooperative state of a node at time t, and i Transition to state j The probability model is as shown in equations (13)-(16):

[0110]

[0111]

[0112] Among them: the transition probability satisfies the normalization, boundedness and symmetry conditions, and the value of the transition probability depends on the possibility of cooperation between satellites due to mission requirements. Specifically, a weighted graph based on the satellite topology structure is constructed to characterize the cooperation relationship between satellites and the matching relationship between satellites and mission requirements.

[0113] In this step, G(V,E) is used to represent the satellite structure topology, where V represents the satellite set and E represents the satellite inter-satellite link (ISL) set. Based on the graph G, a node enhancement graph is proposed: Among them, V st represents the satellite node, V at Represents the subtask node in TK, connecting the edges of structure points It is called a structural edge, which represents the possible ISL set between satellites within the communication range, and the edge connecting the structural point and the attribute point It is called the attribute edge, which represents the matching relationship between the satellite and the subtask. The weights of the structure edge and the attribute edge are ω i,j It indicates that a weight definition is given based on the mission required resources and on-board resources as follows (17):

[0114]

[0115] For structural edges, ω i,jIndicates the similarity of resources between two structural nodes and is positively correlated with the number of common resources between the structural nodes. For attribute edges, when the number of matching resources between the structural node and the attribute node is not 0, ω i,j The value of is positive, ω baseSt and ω baseAt is the correction value of the weight of the structural edge and the attribute edge. The size of the correction value is set according to actual needs. baseSt and ω baseAt The values ​​of are 10 and 100 respectively. In addition, the attribute nodes are not connected to each other;

[0116] In the node-enhanced graph, if multi-hop neighbors Matched with the same subtask, share a new neighbor, the attribute node They become two-hop neighbors, shortening their effective distance in the graph and increasing the possibility of them being classified into the same cluster. The transition probability of structural edges and attribute edges is defined as follows:

[0117]

[0118] Among them, N 1 Represents the number of neighbor structure points of the node, N 2 The number of neighbor attribute points representing the node, by combining equation (17) with equation (18), we get the transition probability matrix between all structure points and attribute points (19):

[0119]

[0120] Step 4.2 Solve the Markov clustering algorithm (MCL) based on the transition probability matrix P

[0121] The MCL algorithm simulates the random walk process through matrix multiplication. The MCL algorithm captures the relationship between related nodes in the clustering process and obtains clusters from the graph structure by expanding and contracting the transition probability matrix P. The MCL algorithm is as follows:

[0122] Step 4.2.1 Expand the transition probability matrix P, that is, P = P ε , simulate the multi-step random walk process to spread the transition probability. Among them, ε is the expansion parameter, which is greater than 1 and represents the number of random walk steps. In order to add inertia to the transition probability and ensure that the transition probability matrix P can finally converge, a self-loop is added to each structural point and attribute node, and the weight of the self-loop is fixed to ω baseAt ;

[0123] Step 4.2.2 performs an expansion operation to further increase the transition probability and suppress the low transition probability. The expansion operation includes performing an r-th power operation on each element of the transition probability matrix P and then normalizing the obtained elements:

[0124]

[0125] Where r>1 represents the expansion parameter, which is used to control the granularity of clustering. The larger the expansion parameter, the more cluster structures will be obtained.

[0126] Step 4.2.3 Repeat steps 4.2.1 and 4.2.2 until the matrix converges and a stable probability distribution is obtained. In the converged transition probability matrix, the high-value elements in each column will be concentrated in the same row as the high-value elements in other columns, indicating that the elements in this row belong to the same cluster.

[0127] Furthermore, in step 5, the probability transfer matrix obtained in step 4 is used to obtain candidate satellite members and candidate cluster head satellites. Based on the cluster head satellite as the center and the constraints in step 3, the function cluster matching the current task is calculated, such as Figure 3 As shown, the specific steps include:

[0128] Step 5.1 Add the task TK′ of selecting cluster head satellite nodes according to the ephemeris information, and construct the node enhancement graph with all satellite nodes as structural points and task node TK′ as the only attribute point. and the probability transformation matrix P′. and P′ are input into the MCL algorithm to obtain the candidate set CH of potential cluster head satellites cd ;

[0129] Step 5.2: Take candidate set CH cd Each cluster head CH i As the center, reduce the maximum hop range hop max The edge weights ω of all satellites other than i,j , and all edge weights ω of the satellites whose subtask requirements have been met i,j ;

[0130] Step 5.3 constructs a new probability transition matrix P, which includes all satellites and all subtasks (except TK′) nodes. Input this matrix into the MCL algorithm to obtain a functional cluster candidate CL i Repeat steps 3 and 4 to obtain a candidate set of functional clusters CL;

[0131] Step 5.4: For each functional cluster candidate CL i Calculate the cluster management cost S(CL) and return the function cluster candidate with the minimum S(CL) as the final function cluster.

[0132] Verification Example

[0133] In the verification of the method of the present invention, a multi-layer low-orbit satellite network is used, which includes a total of 2825 LEO satellites. The various parameters of the satellite constellation are shown in Table 1 below:

[0134] Table 1 Satellite constellation composition

[0135] Satellite layer Layer 1 Layer 2 Layer 3 Layer 4 Number of tracks 6 5 8 32 Number of satellites / orbit 75 75 50 50 Orbital altitude (km) 1325 1275 1130 1110 Orbital inclination (°) 70 81 74 53.8

[0136] The simulation time is 86400s, of which the task generation time is 720s. A snapshot is taken every 60s. By default, the topology of the satellite network remains unchanged during the snapshot. The number of tasks is 5, and the number of subtasks of each task is 5, corresponding to five different on-board payloads. Each satellite has the same processing rate for data packets, which is 5ms. The maximum number of intersatellite links established is 4. In addition, the control information of cluster networking will be transmitted first in the satellite network, so link congestion and link disconnection are not considered. The simulation indicators are: average satellite cluster size, resource waste rate, total hop count and network establishment time.

[0137] The average satellite cluster size refers to the number of all satellites in the satellite cluster, and its mathematical expression is |CL|; the resource redundancy rate refers to the part of the total resources in the satellite cluster that exceeds the mission requirements, and its mathematical expression is shown in formula (9); the total hop count refers to the total number of routing hops from all cluster member satellites to the cluster head satellite in the satellite cluster structure, and its calculation formula is as follows:

[0138]

[0139] The network construction time is divided into two parts: 1) Cluster time T clu , represents the time required to establish communication between the cluster head and the cluster member with the maximum number of hops, 2) the sending time T del , represents the time required to send control information from the control center to the cluster head satellite. Specifically, the control information of the control plane is first uploaded to the possible access satellite in the cluster and then transmitted to the cluster head. It is defined as follows:

[0140] T net =T clu +T del ......(twenty two),

[0141]

[0142] Figure 4a-4cThe influence of the number of different types of task resource requirements on the satellite cluster is indicated, and the algorithm proposed by the method of the present invention is marked as TOC. As the number of task requirements increases, all algorithms show an upward trend, and the performance of the TOC algorithm is always better than other existing algorithms in the field of this technology. In terms of average cluster size, the TOC algorithm reaches 29.6, which is about 40.1% to 43.9% lower than other algorithms; in terms of resource redundancy, the performance of the TOC algorithm is lower than 12.1%, which is about 69.7% lower than other algorithms in the field of this technology. In terms of total hops, the TOC algorithm is about 92 hops, while other algorithms in the field of this technology reach up to 426 hops. This is because other algorithms in the field of this technology often rely on expanding the cluster scale to absorb more nodes in order to meet the task requirements. This expansion of the cluster scale is blind, resulting in inevitable resource redundancy. A satellite that serves a task requirement that has been met may join the satellite cluster due to the expansion of the cluster scale. However, the TOC algorithm solves this problem by actively reducing the weight of the attribute edge of the subtask that has met the requirements.

[0143] Figure 4d The figure shows the influence of different satellite layers on the satellite cluster network construction time. As shown in the figure, the LEACH and LID algorithms both show an upward trend, while the QMM and TOC algorithms show a downward trend. In a single-layer constellation, the network construction time of all algorithms is similar. As the number of constellation layers increases, the network construction time of the TOC algorithm continues to decrease, reaching a minimum value of 36.6ms. This value is nearly 82.8% lower than the LEACH and LID algorithms and 38.2% lower than the QMM algorithm. This is because when there is only one layer of satellites in space, the number of satellites is small and the spatial distribution is sparse. In this case, all clustering algorithms cluster almost all satellites in the mission area to meet the mission requirements, so there is no significant difference between them. However, with the increase of the number of satellite layers and the density of satellites in space, the LID and LEACH algorithms that rely on structure and probability tend to form satellite clusters that are too large, thereby increasing the network construction time. In contrast, the TOC algorithm can find more possible cluster structures by building clusters across layers and reduce the network construction time by controlling the total number of hops.

[0144] In summary, the method of the present invention can effectively realize the clustering networking of satellite networks oriented to mission requirements, reduce cluster size, resource occupancy rate, total number of network hops and network construction time, and improve the on-board resource utilization rate of multi-layer low-orbit satellite giants. Finally, through example verification, the method of the present invention proves its adaptability to multi-layer low-orbit satellite giants.

[0145] The present invention is not limited by the above embodiments. The above embodiments and descriptions are only for explaining the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, which are within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the appended claims.

Claims

1. A task-oriented multi-layer low-orbit satellite giant constellation cluster networking method, comprising the following steps: Step 1. Initialize satellite constellation data, including satellite payload resources, satellite geographic location and orbital information, and satellite inter-satellite link establishment status. When a mission is generated, the mission data is received by the control center located on the ground or in a high-orbit satellite. Step 2. The control center establishes a mission model based on the on-board resources required by the mission, and a satellite model based on the satellite's orbital parameters and on-board payload capacity data; specific mission and satellite model establishment mechanism; Step 3. Determine constraints based on different task types; Step 4. Model the cooperation between satellites and the matching relationship between satellites and tasks as a Markov transition probability model and establish a transition probability matrix; Step 5. Obtain candidate satellite members and candidate cluster head satellites through the transfer probability matrix obtained in step 4, and calculate the functional cluster matching the current task based on the cluster head satellite as the center and the constraints in step 3; Step 6. Upload the networking data to the satellite network. After receiving the networking data, the satellite network establishes a functional cluster through the inter-satellite link status notification mechanism and saves the intra-cluster routing table.

2. According to the task-oriented multi-layer low-orbit satellite giant constellation clustering networking method as described in claim 1, in step 1, the ground or high-orbit satellite control center initializes and maintains various data of the satellite constellation through ephemeris information and satellite information, including the satellite's payload resources, the satellite's geographical location and orbital position information, and the satellite's inter-satellite link establishment status. When a task is generated, the task requirement will first be transmitted to the ground or high-orbit satellite control center, and the control center will execute subsequent steps to divide the satellite task cluster.

3. According to the task-oriented multi-layer low-orbit satellite giant constellation cluster networking method of claim 1, in step 2, the control center establishes a task model according to the on-board resources required by the task, and establishes a satellite model according to the orbital parameters and on-board payload capacity data of the satellite; the specific task and satellite model establishment mechanism is as follows: Step 2.

1. Task model establishment mechanism The ad hoc task is defined as a set TASK, which is composed of the following formula (1): TASK = {task1, task2, …, task TN}...... Equation (1), Among them: task i represents a subtask in an ad hoc task, TN represents the number of subtasks, and the definition of subtask tk is as follows (2)-(3): task i ={type i ,φ i ,time i }......(2), in: type i Indicates the type of subtask, time i represents the execution time required for the subtask, φ i Represents the set of load resources required by the subtask, where: represents the value of the j-th resource requirement of the i-th subtask, RN represents the total number of resource requirements, The larger the value of, the greater the demand of the subtask for this type of resource. If the subtask has no demand for this type of resource, then The value of is 0; Step 2.2 Satellite model establishment mechanism A satellite is defined as the set s i , which is composed of the following formulas (4)-(5): L i ={a,e,θ,Ω,g,v}......(6), N i ={s1,s2,…,s N ,N=|N i |}......(7), Where: id i Indicates the satellite number. Indicates the cluster head satellite number of the cluster to which the satellite belongs, tw i represents the time window of the satellite covering the mission area, δ i Represents the satellite's onboard payload resource set, represents the value of the j-th type of payload resources on the i-th satellite, and One-to-one correspondence, L i Indicates the six orbital numbers of the satellite, a represents the semi-major axis, e represents the eccentricity, θ represents the orbital inclination, Ω represents the longitude of the ascending node, g represents the perigee angular distance, ν represents the true anomaly angle of the perigee epoch, and N i represents the set of neighbor satellites of satellite i, and N represents the number of neighbor satellites.

4. According to the task-oriented multi-layer low-orbit satellite giant constellation cluster networking method of claim 1, in step 3, the restriction conditions are determined according to different task types, specifically: For the functional satellite cluster that meets the mission requirements, the satellite functional cluster overhead formula is proposed as the satellite cluster constraint, and a typical constraint method based on cluster size, resource redundancy rate, total number of hops and mission execution time is given. The constraint conditions are shown in the following equations (8)-(12): S(CL)=α·K+β·R(CL)+γ·H(CL)......(8), CL={CH,CM1,CM2,…,CM K-1 },K=|CL|......(9), Where: S(CL) represents the cluster management cost, CL represents the set of satellite function cluster members, CH represents the cluster head satellite, CM represents the cluster member satellite, K represents the size of the satellite cluster, R(CL) represents the redundancy of resources within the cluster, H(CL) represents the total number of routing hops from the member satellite to the cluster head satellite, α, β, γ are weight parameters used to adjust the impact of various factors on S(CL); Equations (11) and (12) are the resource and time constraints of the satellite function cluster. For a satellite function cluster that meets the mission requirements, the sum of the resources required by all cluster member satellites for a certain mission is greater than the total demand for the resources by the mission, and the time window of any member satellite meets the satellite coverage time requirements of the subtask assigned to the satellite.

5. According to the task-oriented multi-layer low-orbit satellite giant constellation clustering networking method of claim 1, in step 4, the cooperation between satellites and the matching relationship between satellites and tasks are modeled as a Markov transition probability model, and a transition probability matrix is ​​established, which specifically includes the following steps: Step 4.1 Establish a Markov transition probability model The process of satellites being classified into different satellite clusters is regarded as the transfer of the cooperation state between the satellite and different satellites, which is modeled as a Markov process. Let S t represents the set of all possible cooperative states in the satellite network, state represents the specific cooperative state of a node at time t, and i Transition to state j The probability model is as shown in equations (13)-(16): P ij (t)=Pr(S t+1 =state j |S t =state i )......(13), Among them: the transition probability satisfies the normalization, boundedness and symmetry conditions, and the value of the transition probability depends on the possibility of cooperation between satellites due to mission requirements. Specifically, a weighted graph based on the satellite topology structure is constructed to characterize the cooperation relationship between satellites and the matching relationship between satellites and mission requirements. In this step, G(V,E) is used to represent the satellite structure topology, where V represents the satellite set and E represents the satellite inter-satellite link (ISL) set. Based on the graph G, a node enhancement graph is proposed: Among them, V st represents the satellite node, V at Represents the subtask node in TK, connecting the edges of structure points It is called a structural edge, which represents the possible ISL set between satellites within the communication range, and the edge connecting the structural point and the attribute point It is called the attribute edge, which represents the matching relationship between the satellite and the subtask. The weights of the structure edge and the attribute edge are ω i,j It indicates that a weight definition is given based on the mission required resources and on-board resources as follows (17): For structural edges, ω i,j Indicates the similarity of resources between two structural nodes and is positively correlated with the number of common resources between the structural nodes. For attribute edges, when the number of matching resources between the structural node and the attribute node is not 0, ω i,j The value of is positive, ω baseSt and ω baseAt is the correction value of the weight of the structural edge and the attribute edge. The size of the correction value is set according to actual needs. baseSt and ω baseAt The values ​​of are 10 and 100 respectively. In addition, the attribute nodes are not connected to each other; In the node-enhanced graph, if multi-hop neighbors and Matched with the same subtask, share a new neighbor, the attribute node They become two-hop neighbors, shortening their effective distance in the graph and increasing the possibility of them being classified into the same cluster. The transition probability of structural edges and attribute edges is defined as follows: Among them, N1 represents the number of neighboring structural points of the node, and N2 represents the number of neighboring attribute points of the node. By combining formula (17) with formula (18), the transition probability matrix between all structural points and attribute points is obtained: Step 4.2 Solve the Markov clustering algorithm (MCL) based on the transition probability matrix P The MCL algorithm simulates the random walk process through matrix multiplication. The MCL algorithm captures the relationship between related nodes in the clustering process and obtains clusters from the graph structure by expanding and contracting the transition probability matrix P. The MCL algorithm is as follows: Step 4.2.1 The transfer probability matrix P is expanded, that is, P = P ε , simulate the multi-step random walk process to spread the transition probability. Among them, ε is the expansion parameter, which is greater than 1 and represents the number of random walk steps. In order to add inertia to the transition probability and ensure that the transition probability matrix P can finally converge, a self-loop is added to each structural point and attribute node, and the weight of the self-loop is fixed to ω baseAt ; Step 4.2.2 performs an expansion operation to further increase the transition probability and suppress the low transition probability. The expansion operation includes performing an r-th power operation on each element of the transition probability matrix P and then normalizing the obtained elements: Where r>1 represents the expansion parameter, which is used to control the granularity of clustering. The larger the expansion parameter, the more cluster structures will be obtained. Step 4.2.3 Repeat steps 4.2.1 and 4.2.2 until the matrix converges and obtains a stable probability distribution. In the converged transfer probability matrix, the high-value elements in each column will be concentrated in the same row as the high-value elements in other columns, indicating that the elements in this row belong to the same cluster.

6. According to the task-oriented multi-layer low-orbit satellite giant constellation clustering networking method of claim 1, step 5, through the probability transfer matrix obtained in step 4, obtain candidate satellite members and candidate cluster head satellites, take the cluster head satellite as the center and the restriction conditions in step 3 as the basis, calculate the function cluster matching the current task, specifically including the following steps: Step 5.1 Add the task TK′ of selecting cluster head satellite nodes according to the ephemeris information, and construct the node enhancement graph with all satellite nodes as structural points and task node TK′ as the only attribute point. and the probability transformation matrix P'. and P′ are input into the MCL algorithm to obtain the candidate set CH of potential cluster head satellites cd ; Step 5.2: Take candidate set CH cd Each cluster head CH i As the center, reduce the maximum hop range hop max The edge weights ω of all satellites other than i,j , and all edge weights ω of the satellites whose subtask requirements have been met i,j ; Step 5.3 constructs a new probability transition matrix P, which includes all satellites and all subtasks (except TK′) nodes. Input this matrix into the MCL algorithm to obtain a functional cluster candidate CL i Repeat steps 3 and 4 to obtain a candidate set of functional clusters CL; Step 5.4: For each functional cluster candidate CL i Calculate the cluster management cost S(CL) and return the function cluster candidate with the minimum S(CL) as the final function cluster.

Images