Large-scale satellite group network self-organizing clustering method based on alliance composition game

Abstract

The invention relates to a large-scale satellite group network self-organizing clustering method, which comprises the following steps of: acquiring position data of all satellites in a ground station server, identifying different satellites on the same layer, and establishing a large-scale satellite group-oriented self-organizing clustering mathematical model and a large-scale satellite group clustering model construction mode based on an alliance composition game; and finally, forming a self-organizing clustering algorithm based on a coalition game. And finally, effective dimension reduction of the large-scale satellite group online collaboration problem is realized.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent cluster self-organizing control, and particularly relates to a self-organizing clustering method for a large-scale satellite constellation network. Background Art

[0002] With the development of new technologies such as microelectronics, micro-electromechanics, and integrated circuits, and the innovation of satellite design and development concepts, satellites have higher technical performance, lower economic costs, and shorter development cycles. Large-scale satellite systems have entered a period of rapid development. A large-scale satellite constellation system generally consists of hundreds, thousands, or even tens of thousands of satellites operating in different orbital planes, altitudes, and phases, which can greatly enrich the support of remote sensing data and space-based information service means, and is of great significance for China to build a strong space power, safeguard national defense security, and develop social economy.

[0003] By regarding each satellite in the satellite system as an intelligent agent for autonomous decision-making, large-scale satellite constellation online collaboration means that each satellite adjusts its own decisions and behaviors online to achieve specific optimization goals or maintain the consistency of the overall system behavior, ensuring the maximization of the overall system efficiency. On this basis, the large-scale satellite constellation collaboration architecture represents the organizational form when each satellite in the satellite system works, clarifies the decisions and functions of each satellite, and further clarifies the organizational relationship and data flow between each satellite on this basis. Further, considering the large-scale characteristics of the large-scale satellite constellation, most large-scale satellite constellation collaboration architecture studies draw on the collaboration architecture of swarm intelligence systems and adopt a clustering strategy of "divide and conquer". This strategy provides an effective and feasible solution for the management of large-scale satellite systems through regional division and hierarchical autonomy.

[0004] As Figure 1 shown, a large-scale satellite constellation network system containing multiple heterogeneous and different orbital altitude constellations can be generalized to a Mobile Ad-hoc NETworks (MANET). UAV networks, vehicle networks, and radio networks, etc. are all typical examples of MANET. The core difference of these networks lies in the communication link characteristics between nodes. Different from these networks above, the connectivity between any two satellites in the large-scale satellite constellation network depends on the visibility relationship and effective line-of-sight range between the satellites. As Figure 2 shown, considering satellite i and satellite j in different orbital planes, when the two satellite nodes are within the effective line-of-sight range and the geocentric angle formed by the two is less than the maximum visible geocentric angle minus the allowable angle offset caused by perturbation, a feasible communication link can be established. Among them, di,j is the distance between satellite i and satellite j, which can be calculated by the Euclidean distance of the spatial coordinates of two points in the J2000.0 geocentric inertial coordinate system, and θ i,j and respectively represent the geocentric angle and the maximum visible geocentric angle between satellite i and satellite j.

[0005] Inter-satellite communication mainly includes two methods: inter-satellite laser link communication and inter-satellite microwave communication. There are differences in the communication methods between satellites at different orbital altitudes, as Figure 3 shown. Although the inter-satellite laser link has significant advantages in communication capabilities and measurement accuracy, it is limited by the performance of the current acquisition, tracking, and pointing (ATP) system and cannot meet the requirement of direct communication with multiple satellites simultaneously. Therefore, it cannot completely replace the existing microwave-based concurrent time-division and space-division inter-satellite link system. To ensure the autonomous cooperation performance of the satellite constellation network, satellites need to frequently perform inter-satellite measurements with multiple satellites in a short time. Therefore, according to the link establishment rules of the satellite system, the present invention adopts Figure 3 the composite laser-microwave link scheme shown in (c). Specifically, inter-satellite laser communication links are used within the constellation at the same orbital altitude, while microwave communication is used between different orbital altitudes.

[0006] Based on the above background technology of satellite constellations and the characteristics of inter-satellite link connections in satellite networks, the self-organizing clustering problem of large-scale satellite constellations faces the following problems: 1) Existing clustering models often ignore the communication and connection characteristics of inter-satellite links during construction and usually assume that inter-satellite links are always connected in real time. This assumption may hold in small-scale satellite constellation scenarios, but it is difficult to meet in large-scale satellite constellation systems. 2) Existing satellite self-organizing clustering algorithms are designed for a limited number of satellite constellations and have limited applicability in the environment of dynamic changes in on-board networks. 3) Most existing on-board self-organizing clustering algorithms are heuristic algorithm methods designed based on problem characteristics and lack theoretical support and performance guarantees. In summary, considering the realistic factors such as heterogeneous nodes and inter-satellite links and frequent network topology switching in large-scale satellite constellations, there is an urgent need for a decentralized self-organizing clustering algorithm for satellite constellation networks with theoretical support to effectively reduce the dimension of the online cooperation problem of large-scale satellite constellations. Summary of the Invention

[0007] To address the problems of the above existing technologies, the present invention provides a method for self-organizing clustering of large-scale satellite constellation networks. Aiming at the problems that the self-organizing clustering of large-scale satellite constellation networks lacks a standardized mathematical model and a decentralized self-organizing clustering algorithm with theoretical support, 1) a self-organizing clustering mathematical model for large-scale satellite constellations is established, 2) a method for constructing a large-scale satellite constellation model based on coalition formation game is designed, and 3) a self-organizing clustering method based on coalition formation game is provided.

[0008] The technical solution adopted by the present invention is as follows:

[0009] A method for self-organizing clustering of large-scale satellite constellation networks, which performs the following steps:

[0010] S1. Obtain the position data of all satellites in the ground station server and identify different satellites at the same layer;

[0011] S2. Establish a self-organizing clustering mathematical model for large-scale satellite constellations;

[0012] S21. Perform modeling of the satellite constellation network system;

[0013] S22. Determine the self-organizing clustering revenue function for large-scale satellite constellations;

[0014] S23. Form a self-organizing clustering mathematical model for large-scale satellite constellations;

[0015] S3. Construction method of a large-scale satellite constellation clustering model based on coalition formation game;

[0016] S31. Based on the coalition formation game model;

[0017] S32. Determine the coalition formation rules;

[0018] S33. Determine the preference order of the participants;

[0019] S4. Self-organizing clustering algorithm based on coalition formation game;

[0020] S41. Construct an asynchronous decision-making algorithm framework;

[0021] S42. Form a self-organizing clustering algorithm A1;

[0022] S43. Form a self-organizing clustering algorithm A2.

[0023] According to whether the current cluster to which node i belongs satisfies the CFPLS problem constraint, the designed clustering algorithm A1 and clustering algorithm A2 are respectively executed;

[0024] If the current cluster to which node i belongs satisfies the problem constraint in CFPLS, that is, if and only if the cluster C to which node i belongs m satisfies

[0025]

[0026] then node i executes clustering algorithm A1, otherwise it executes clustering algorithm A2;

[0027] Among them, in the said problem constraint, N max and D max are respectively the maximum limits of the cluster size and the cluster diameter.

[0028] Furthermore, in step S21,

[0029] For different communication links, the communication methods adopted are different;

[0030] There are differences in the transmission rates of inter-satellite transmissions.

[0031] Furthermore, in step S21,

[0032] The orbital radii of satellites i and j in the same layer are the same, that is

[0033] The link transmit power P t , the maximum directive antenna gain G 0 and the channel power gain h i,j , the inter-satellite link transmission rate rt i,j is:

[0034]

[0035] where K and T represent the Boltzmann constant and the total system noise temperature respectively, represents the ratio requirement of the received energy per bit to the noise power, and M represents the link margin.

[0036] Furthermore, in step S21,

[0037] L(d i,j ) represents the distance-dependent free space propagation loss, where d i，j is the distance between satellites i and j. Given the speed of light c and the communication electromagnetic wave frequency F, its calculation formula is as follows:

[0038]

[0039] The orbital radii of satellites i and j in different layers are different, that is

[0040] Furthermore, in step S21,

[0041] For the channel bandwidth B of the inter-satellite link, the achievable transmission rate rt is calculated using the Shannon formula i,j , where rt i,j is a parameter:

[0042] rt i,j = B log(1 + Λ i,j )

[0043]

[0044] where Λ i,j represents the signal-to-noise ratio SINR of the inter-satellite link between satellites i and j.

[0045] Furthermore, in step S21,

[0046] The inter-satellite link transmission rate rt between any two satellites i and j is expressed as: i,j is:

[0047]

[0048] Further, in step S21,

[0049] When a direct available inter-satellite link cannot be established between satellite i and satellite j, a multi-hop inter-satellite link is established through other satellite nodes as relays; the link capacity of the multi-hop inter-satellite link is calculated based on all the inter-satellite links passed through;

[0050] Denote the shortest path between satellite i and satellite j as q i,j = ((i, i 1 , (i 1 , i 2 ,...,(i h , j)), then the link capacity of the multi-hop inter-satellite link between satellite i and satellite j is:

[0051]

[0052] where α is the loss coefficient of the multi-hop path, usually taking α < 1. The more the number of path jumps, the more the link capacity is lost; when |q i,j | = 1, it means that there is no need for transit between nodes. At this time, κ i,j = rt i,j , and κ i,j represents the link capacity of the inter-satellite link between any two satellites in the satellite constellation network.

[0053] Further, in step S21,

[0054] The satellite constellation network is represented as an undirected graph (V, E), where V represents the set of satellite nodes and E represents the set of edges connecting the satellites; when l i,j = 1, (i, j) ∈ E, satellite i and satellite j can be regarded as neighbor satellites to each other. The inter-satellite link relationship of the large-scale satellite constellation network can be represented by the adjacency matrix l:

[0055]

[0056] Further, in step S21,

[0057] For each edge in E, the capacity value of the corresponding inter-satellite link can be calculated;

[0058] The inter-satellite link capacity of the large-scale satellite constellation network is represented by the link capacity matrix κ:

[0059]

[0060] Further, in step S22,

[0061] In the cluster structure Π of a large-scale constellation network, the collaboration ability χ total is expressed as:

[0062]

[0063] where χ inter (C m ) and χ intra (C m ) represent the inter-cluster collaboration ability and intra-cluster collaboration ability of cluster C m respectively; the inter-cluster collaboration ability χ inter (C m ) is related to the link ability of the inter-satellite link between the cluster head node, and j ∈ (V Head ∩C m ) represents the cluster head node of cluster C m , V Head represents the set of cluster head nodes in the cluster structure, and β inter is a parameter for adjusting the importance of the inter-cluster collaboration ability, representing the proportion of the inter-cluster collaboration ability in the total collaboration ability of the cluster structure; the intra-cluster collaboration ability χ intra (C m ) is related to the link ability between intra-cluster members. When the inter-satellite links between intra-cluster members all have stronger link abilities, the intra-cluster collaboration ability is also stronger. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0065] Figure 1 is an example diagram of the connection of inter-satellite links in a large-scale constellation system.

[0066] Figure 2 is an example diagram of the visibility of satellites in different orbital planes.

[0067] Figure 3 is a schematic diagram of the inter-satellite communication link between different orbital altitudes.

[0068] Figure 4 is a schematic diagram of cluster head election operation, bilateral exchange operation, and improved replacement operation.

[0069] Figure 5 is a schematic diagram of the asynchronous decision algorithm framework.

[0070] Figure 6It is the change in the number of different types of nodes and the number of times operations are selected during the coalition formation process. Detailed implementation manners

[0071] The following describes exemplary embodiments of the present application with reference to the accompanying drawings. Various details of the embodiments of the present application are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, descriptions of well-known functions and structures are omitted below for clarity and conciseness.

[0072] To enable those skilled in the art of this technology to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0073] S1. Obtain the position data of all satellites in the ground station server and identify different satellites in the same layer;

[0074] S2. Establish a self-organizing clustering mathematical model for large-scale satellite clusters;

[0075] S21. Modeling of the satellite cluster network system

[0076] Based on the inter-satellite link connection characteristics of the satellite network, different communication methods are adopted for different communication links, resulting in differences in the transmission rates of inter-satellite transmissions. The present invention will introduce the channel models between satellites in the same layer and different layers to characterize the impact of channel changes on the transmission rate between satellites.

[0077] For satellites i and j in the same layer, their orbital radii are the same (i.e., ), under the conditions of the given link transmit power P t , the maximum directive antenna gain G 0 and the channel power gain h i,j , according to the above analysis, in practical engineering applications, the transmission rate rt i,j between satellites in the same layer can be expressed by the following formula

[174] :

[0078]

[0079] Where K and T represent the Boltzmann constant and the total system noise temperature respectively, It represents the requirement for the ratio of received energy per bit to noise power, and M represents the link margin.

[0080] L(d i,j ) represents the distance - related free - space propagation loss, where d i，j is the distance between satellite i and satellite j. When the speed of light c and the communication electromagnetic wave frequency F are given, its calculation formula is as follows:

[0081]

[0082] For satellites i and j in different layers, their orbital radii are different (i.e., ), given the channel bandwidth B of the inter - satellite link, the achievable transmission rate rt

[175] can be calculated using Shannon's formula i,j , where rt i,j is a parameter:

[0083] rt i,j = B log(1 + Λ i,j ) (3)

[0084]

[0085] where Λ i,j represents the signal - to - interference - plus - noise ratio (SINR) of the inter - satellite link between satellite i and satellite j.

[0086] Based on the above analysis, given the relevant parameters of the communication link, the link transmission rate rt i,j of the same - layer satellites and different - layer satellites can be obtained by combining the above formula. Therefore, in the entire satellite constellation network, the inter - satellite link transmission rate rt i,j between any two satellites i and j can be expressed by the following formula:

[0087]

[0088] When a direct available inter - satellite link cannot be established between satellite i and satellite j, it may be necessary to use other satellite nodes as relays to establish a multi - hop inter - satellite link. At this time, the link capacity of the multi - hop inter - satellite link needs to be calculated based on all the inter - satellite links it passes through. Denote the shortest path between satellite i and satellite j as q i,j = ((i,i 1 ),(i 1 ,i 2 ),...,(i h ,j)), then the link capacity of the multi - hop inter - satellite link between satellite i and satellite j can be calculated by the following formula:

[0089]

[0090] Among them, α is the loss coefficient of the multi-hop path. Usually, α < 1. The more the number of path jumps, the more the link capacity loss. When |q i,j | = 1, it means that there is no need for transit in the link between nodes. At this time, κ i,j = rt i,j . Therefore, κ i,j can be used to represent the link capacity of the inter-satellite link between any two satellites in the constellation network.

[0091] Consider representing the constellation network as an undirected graph (V, E), where V represents the set of satellite nodes and E represents the set of edges connecting the satellites. When l i,j = 1, (i, j) ∈ E, and satellite i and satellite j can be regarded as neighbor satellites to each other. The inter-satellite link relationship of the large-scale constellation network can be represented by the adjacency matrix l:

[0092]

[0093] For each edge in E, the capacity value of the corresponding inter-satellite link can be calculated according to formula (7). Therefore, the inter-satellite link capacity of the large-scale constellation network can also be represented by the link capacity matrix κ:

[0094]

[0095] S22. Determine the self-organizing clustering revenue function for large-scale constellations;

[0096] The problem of large-scale constellation clustering aims to serve the online collaboration of satellites within and between clusters. Therefore, a strong cluster structure collaboration ability is the key to achieving efficient hierarchical management and control. At the same time, since the satellites are in a continuous motion state, the large-scale constellation network constitutes a dynamic network, and the maintenance and update of its cluster structure will bring certain management overhead. In addition, in order to achieve the collaborative effect, satellite nodes need to communicate and transmit data, which will also generate additional management overhead. Based on this, the present invention will first model the collaboration ability and management overhead of the cluster structure, and then establish the revenue function of the CFPLS problem based on these two metrics.

[0097] Cluster collaboration ability

[0098] In a constellation network, the cooperation among satellite nodes mainly depends on the information interaction and transmission of inter-satellite links. No matter what tasks the constellation network cooperatively executes or what services it provides, the system needs to ensure stable and efficient data transmission among satellite nodes. The transmission rate of inter-satellite links is a key factor affecting the constellation cooperation ability. Therefore, referring to Massin's definition of the cooperation ability of cluster structures, in this section, the link transmission ability of inter-satellite communication links is used to quantify the cooperation ability of cluster structures. For the cluster structure Π of a large-scale constellation network, its cooperation ability χ total usually needs to be considered from two aspects, as shown in the following formula:

[0099]

[0100]

[0101] where χ inter (C m ) and χ intra (C m ) represent the inter-cluster cooperation ability and intra-cluster cooperation ability of cluster C m respectively; the inter-cluster cooperation ability χ inter (C m ) is related to the link ability of the inter-satellite link between the cluster head node and the cluster, j ∈ (V Head ∩ C m ) represents the cluster head node of cluster C m , V Head represents the set of cluster head nodes in the cluster structure, β inter is a parameter to adjust the importance of the inter-cluster cooperation ability, representing the proportion of the inter-cluster cooperation ability in the total cooperation ability of the cluster structure; the intra-cluster cooperation ability χ intra (C m ) is related to the link ability among intra-cluster members. When the inter-satellite links among intra-cluster members all have strong link abilities, the intra-cluster cooperation ability is also strong. l i,j ∈ {0, 1} and κ i,j ∈ R + represent the visibility relationship and link transmission ability between nodes respectively. On the premise of given communication-related parameters, the link transmission ability κ i,j is inversely proportional to the distance between two nodes.

[0102] The cluster structure is the final clustering scheme composed of multiple clusters. Taking the nodes {1, 2, 3, 4, 5, 6} as an example, the cluster structure is {{1, 3, 2}, {4}, {5, 6}}, where cluster Cm is one of the clusters, such as {1, 3, 2}.

[0103] Cluster management overhead

[0104] The present invention uses the communication overhead generated by maintaining the cluster structure as a measure of the cluster structure management overhead. Similar to the modeling of the collaborative ability of the cluster structure, the cluster structure management overhead consists of two parts: the inter-cluster management overhead and the intra-cluster management overhead. The calculation formula is as follows:

[0105]

[0106] Among them, E inter (C m ) and E intra (C m ) are the inter-cluster management overhead and the intra-cluster management overhead of cluster C m , respectively.

[0107] The calculation of the inter-cluster management overhead is based on the cluster head node broadcasting data packets containing local topology information and task information to other cluster heads. Let the average size of the inter-cluster data packet be V inter , then the inter-cluster management overhead of cluster C m is:

[0108] E inter (C m ) = ∑V inter = V inter (|V Head | - 1) (13)

[0109] Among them, represents the cluster head node of cluster C m , and |V Head | represents the size of the set of cluster head nodes.

[0110] The intra-cluster management overhead involves each member in the cluster broadcasting data packets containing local status information and task information to other members. Let the average size of this data packet be V intra , then the intra-cluster management overhead of cluster C m is:

[0111]

[0112] When constructing a large-scale star cluster structure, it is necessary to balance the collaborative ability and management overhead of the cluster structure. Based on the collaborative ability model and management overhead model introduced in this section, a benefit function for the large-scale star cluster structure can be established, and the expression is as follows:

[0113]

[0114] Among them, the weight factor ξ ∈ [0, 1] is used to weigh the influence of the collaborative ability and management overhead indicators on the benefit function, and are the normalized indicators respectively, χ max and E maxRespectively represent the theoretical maximum values of the cluster structure collaboration ability and management overhead.

[0115] S23. Form a self-organizing clustering mathematical model for large-scale satellite constellations;

[0116] For a large-scale satellite constellation, all satellite nodes in it need to be assigned to several mutually exclusive clusters. Under the constraints of cluster scale and cluster diameter, find the optimal cluster structure Π = {C 1 , C 2 ,..., C Π}, and maximize the profit function G(Π), whose expression is as follows:

[0117]

[0118] Where N max and D max are the maximum limits of the cluster scale and cluster diameter respectively.

[0119] Above, a self-organizing clustering model for large-scale satellite constellations is established, and the mathematical expressions of decision variables, constraint conditions, and objective functions in the model are given in a standardized way, which helps the management and control department understand the combinatorial optimization essence of the self-organizing clustering problem of large-scale satellite constellations, and then better guide the design of algorithms and operators.

[0120] S3. Construction method of a large-scale satellite constellation clustering model based on coalition formation game;

[0121] Based on the above model, in this section, a self-organizing clustering framework for large-scale satellite constellation networks is designed based on the coalition formation game theory. As an important branch of game theory, the coalition formation game explores the action strategies of cooperative behaviors among participants, thus providing a theoretical analysis framework for solving the satellite constellation clustering problem in a decentralized decision-making manner. Specifically, by designing the coalition value function, coalition formation rules, and participant preference order in the coalition formation game, the global objective function in the satellite constellation clustering problem is decomposed into local utility functions of individual satellites in this section, providing a basis for the operation operators of individual decision-making in subsequent self-organizing clustering methods. S31. Based on the coalition formation game model;

[0122] The large-scale satellite constellation clustering model based on the coalition formation game is as follows:

[0123] G C = <V, {a i} i∈V , V> (20)

[0124] V = {1, 2,..., |V|} is the set of participants, corresponding to the set of satellites in the CFPLS problem;

[0125] is the set of actionable actions for all participants \(i\) (i.e., satellite \(i\)), where the symbol represents the Cartesian product, and \(c\) i represents the set of potential coalitions that satellite \(i\) can choose to join, and \(h\) i represents whether the current node is to become the cluster head satellite. \(h\) i = 1 means that satellite \(i\) serves as the cluster head satellite, and conversely \(h\) i = 0 means that satellite \(i\) is an ordinary member satellite.

[0126] \(V(C\) m ) is the coalition value function, representing the total value generated by coalition \(C\) m in the coalition structure \(\Pi\), and its definition is as follows:

[0127]

[0128] where \(\xi\) is a weight factor used to weigh the impact of the collaboration ability and management overhead metrics on the revenue function, and \(\chi\) total (\(C\) m ) and \(E\) total (\(C\) m ) respectively represent the sum of the collaboration ability and management overhead in cluster \(C\) m , and \(\chi\) max and \(E\) max respectively represent the theoretical extrema of the cluster structure's collaboration ability and management overhead. Given the premise of a feasible coalition structure \(\Pi=\{C\) 1 ,..., \(C\) |Π| \}, the sum of the value functions of all coalitions constitutes the objective function \(G(\Pi)\) of the CFPLS problem, and its derivation is as follows:

[0129]

[0130] where, and are the normalized collaboration ability and management overhead respectively, and satisfy the constraint conditions of the cluster size and cluster diameter.

[0131] S32. Determine the coalition formation rules;

[0132] Based on the above coalition formation game model, this study further designs a series of coalition formation rules among satellite nodes.

[0133] Definition 1 Cluster head election operation: For any node \(i\), its cluster head election operation is defined as the cluster head of the cluster \(C\) k to which node \(i\) belongs changes from the original to node \(i\), that is

[0134]

[0135] Definition 2 Bilateral exchange operation: For any node i, the bilateral exchange operation is defined as the coalition C k to which node i belongs is converted to another coalition C l ∈Π (l ≠ k), and at the same time node i' is converted from the coalition C l to the coalition C k to which node i belongs, that is

[0136]

[0137] Definition 3 Improved replacement operation: For any node i that forms a separate cluster, the improved replacement operation is defined as node i joining the coalition C l ∈Π and replacing node i' in C l At the same time, node i' is converted to another coalition C p ∈Π (l ≠ p), that is

[0138]

[0139] The addition of the above three new operations to the coalition formation rules provides three significant advantages. First, the cluster head election operation allows nodes to elect cluster heads during the coalition formation process, which helps to achieve unified modeling and solution of the CFPLS problem. Second, the bilateral exchange operation and the improved replacement operation provide the possibility of flexible adjustment when the number of coalition members is full, in order to pursue higher coalition division benefits. Finally, both the bilateral exchange operation and the improved replacement operation are equivalent to the combination of two coalition conversion operations, which can effectively accelerate the coalition formation speed and improve the self-organizing clustering efficiency. To more intuitively understand the above three new coalition operations, Figure 4 shows the schematic diagram of the coalition structure before and after these three operations.

[0140] S33. Determine the participant preference order

[0141] Within the framework of the coalition formation game, each participant has the right to join any coalition that meets the established constraints. Given that participants obtain different benefits from different coalitions, they have different preferences for joining different coalitions. This concept is called the participant preference order in the coalition formation game

[177] . Before constructing the participant preference order, it is necessary to first quantify the advantages and disadvantages of each node's operation under the coalition formation rules. This section proposes the concepts of operation benefit and operation gain for this purpose.

[0142] Definition 4 Operation benefit: For any operation σ(i) in node i, the arbitrary operation refers to any way of data processing of satellite nodes. After the operation, its coalition set is changed to then the operation benefit r(σ(i)) of this operation is defined as the coalition set The sum of the coalition value functions of all coalitions in, that is

[0143]

[0144] Definition 5 Operational gain: For the coalition set involved in any operation σ(i) in node i After the operation, its coalition set changes to Then the operational gain g(σ(i)) of operation σ at this time is the difference in the coalition value functions of the coalition sets involved before and after node i performs the operation. The specific expression is as follows:

[0145]

[0146] Among them, V(C m ) is the coalition value function of coalition C m , and its specific meaning has been detailed in formula (3-25) above.

[0147] According to the above definition of operational benefits and combined with the definition of specific operations in the coalition formation rules, the specific operational benefit values of each operation of node i can be obtained. After obtaining these operational benefit values, the participant preference order can be defined in the following way:

[0148] Definition 6 Participant preference order: The participant preference order Can be defined as a complete and transitive binary relation between two feasible coalition operations σ 1 (i) and σ 2 (i) of node i. When the operational benefit of coalition operation σ 1 (i) is greater than the operational benefit of σ 2 (i), then it is called Called a strong preference relation, that is

[0149]

[0150] Similarly, a weak preference relation Can be defined as follows

[0151]

[0152] S4. Self-organizing clustering algorithm based on coalition formation game;

[0153] Based on the coalition formation game theory, a self-organizing clustering framework suitable for large-scale satellite constellation networks is designed. By constructing corresponding coalition formation rules, the global objective function of the satellite constellation clustering problem is effectively decomposed into the preference orders of individual satellite nodes. Based on this, satellite nodes can update their clustering actions according to their respective coalition formation rules and preference orders. The present invention further proposes a general asynchronous online clustering algorithm: the Coalition formation game-based Self-Organized Clustering (CSOC) algorithm.

[0154] S41. Construct an asynchronous decision algorithm framework;

[0155] The CSOC algorithm adopts a self-organizing asynchronous decision framework to adapt to the dynamics of satellite nodes in the self-organizing clustering problem of satellite constellations. Figure 5 The self-organizing asynchronous decision framework presented in the form of a finite state machine model of nodes is shown, which constitutes the core structure of the CSOC algorithm.

[0156] As Figure 5 shown, node i exists in two states: the idle state and the busy state.

[0157] Idle state: When node i does not receive information sent by other nodes to change its own node cluster structure, it will remain in the idle state for a random time Ti. When node i receives information from other nodes, it modifies its belonging cluster according to the information and re-evaluates whether it has maintained the idle state for more than Ti. If it has, it checks whether the cluster structure state has changed; if it has changed, it transfers to the busy state; otherwise, it continues to remain in the idle state.

[0158] Busy state: It means that node i is running a clustering algorithm to determine its new belonging cluster and the cluster structure of associated nodes. According to whether the cluster that node i currently belongs to satisfies the constraints of the satellite constellation clustering problem, the designed clustering algorithms A1 and A2 are respectively executed. After the algorithm is executed, the clustering of the node is updated according to the result, and the relevant nodes are notified of the change information of their belonging clusters.

[0159] Since the topology of the satellite constellation network may change over time, a cluster that satisfied the constraints at a past time may no longer satisfy the constraints at the current time. Therefore, when Ti ends, the node will check whether the CFPLS problem constraints of its current cluster are satisfied. If so, algorithm A1 is executed to solve it. If there are no constraints, the node will execute algorithm A2 to solve the clustering problem when the constraints are violated.

[0160] S42. Form a self-organizing clustering algorithm A1;

[0161] The basic idea of Algorithm A1 is as follows: Based on meeting the constraint conditions, Algorithm A1 first explores all coalition transformation operations, improvement replacement operations, and bilateral exchange operations with positive operation gains to form a set of optional operations. If this set is non-empty, then according to the participant preference order, the coalition operation with the maximum operation gain is selected as the key operation. Subsequently, according to the action update strategy, it is decided whether to update the cluster to which it belongs based on the current operation. If it is decided to update, the cluster structure change information is sent to the relevant nodes. Finally, if the current cluster is successfully updated, it is decided whether to replace the cluster head of the updated cluster based on the value of the cluster head election operation; if it is not updated, the replacement of the cluster head is directly checked and decided. The pseudocode of Algorithm A1 is shown as follows.

[0162]

[0163]

[0164]

[0165] In Algorithm A1, a feasible operation set is found and the important operation σ is obtained according to the participant preference order. max (i) After (as shown in line 15 of Algorithm A1), two operation update strategies are set: namely, the Greedy strategy and the Log-linear learning strategy. Let Pr(σ max (i)) represent the probability of updating the cluster structure of node i using the important operation σ max (i).

[0166] 1) The greedy strategy pursues the maximum immediate benefit for each step of decision-making and directly uses σ max (i) as the update operation, that is, Pr(σ max (i)) = 1;

[0167] 2) The action update strategy based on log-linear learning uses the operation benefit as an index to measure whether to use the important operation σ max (i) to update the cluster structure, and its operation update probability expression is as follows:

[0168]

[0169] In the formula, ε indicates the possibility of the node taking a suboptimal behavior. When ε → 0, the node will select to update the important operation with a high probability; conversely, the node will retain the current state or update σ max (i) with equal probability.

[0170] S43. Form the self-organizing clustering algorithm A2;

[0171] The basic idea of Algorithm A2 is the same as that of Algorithm A1, which is to sequentially find feasible operations in the neighbor clusters and update the maximum revenue operation. The differences between the two are as follows: 1) The initialization of the set of feasible operations P i is different. Algorithm A2 is designed for nodes in the current cluster that do not satisfy the constraints. Therefore, during initialization, the node will choose to leave the current cluster and form a single-node cluster; 2) The operation evaluation metrics are different. Algorithm A1 selects operations based on whether the operation gain is positive, while in Algorithm A2, since it is oriented towards constraint violation situations, the operation gains of all feasible operations are positive infinity. Therefore, the operation revenue is used as the evaluation metric. 3) The implementation results are different. In A1, the number of cluster structures always remains unchanged or decreases. In A2, since it is possible for node i to separate from the current cluster C k to satisfy the problem constraints, the number of clusters in the cluster structure may increase. 4) The operation selection strategies are different. In Algorithm A1, greedy strategies and log-linear learning strategies can be used for operation updates, while in Algorithm A2, only the greedy strategy is used for operation updates.

[0172] The specific pseudocode of Algorithm A2 is as follows:

[0173]

[0174]

[0175] To verify the effectiveness of the proposed clustering method, the present invention selects four typical two-layer constellation scenarios as the input conditions for the large-scale satellite constellation clustering method. Without loss of generality, each scenario consists of an upper-layer Walker constellation and a lower-layer Walker constellation. The specific parameter settings are shown in the following table.

[0176] Table 1 Basic Information of Large-Scale Satellite Constellation Network Scenarios

[0177]

[0178] In the multi-layer heterogeneous constellation of the present invention, the connection situation of the inter-satellite links is set to meet the actual situation of the link connection. Specifically, satellites in the same layer are connected by laser links, while satellites in different layers are connected by microwave links. The effective line-of-sight range D LoS between satellites is set to the maximum distance at which two satellites in the scenario can see each other. Regarding the parameter settings related to the links, based on the existing technology, the link transmission capacity between satellites in the same layer is between 20 Mbps and 100 Mbps, while the link transmission capacity between satellites in different layers is between 1 Mbps and 20 Mbps, and the transmission capacity is negatively correlated with the inter-satellite distance.

[0179] Given that the relevant research on the problem of large-scale satellite constellation clustering is relatively limited, the present invention refers to the experimental settings of recent research and sets the network topology data sampling interval of the satellite constellation network to 100 seconds. In addition, considering that the present invention mainly focuses on the performance of the clustering scheme rather than the routing scheme, without loss of generality, α, β inter , V inter and V intra in the model are respectively set to 0.8, 2, 5 and 0.5. The weight factor ξ ∈ [0, 1] is used to balance the influence of the cooperation ability and management overhead metrics on the revenue function, which is generally determined by expert judgment and is set to 0.3 in this experiment. Unless otherwise specified, the maximum cluster size N max and the maximum cluster diameter D max are respectively defaulted to and 2, where |V| represents the total number of nodes.

[0180] First, by running the CSOC-G algorithm on Scenario 4 to form a cluster structure and record the corresponding parameter indicators, the process of forming a cluster structure in the satellite constellation network is analyzed. The corresponding parameter indicators include the number of isolated nodes, cluster head nodes and member nodes in the network.

[0181] The meaning of an isolated node is a node that forms a single cluster in the network.

[0182] A cluster head node refers to a node that serves as a cluster head in the network. According to the problem setting, there is only one cluster head in a cluster, and the number of cluster head nodes is the same as the number of clusters in the cluster structure.

[0183] A member node refers to other nodes in the cluster that do not serve as cluster head nodes.

[0184] Taking the satellite constellation network in Scenario 4 at the start time of the simulation as an example, the proposed self-organizing clustering algorithm is used for clustering and the satellite network state and algorithm-related parameters of each iteration are recorded to analyze the process of forming a star cluster structure. Figure 6 shows the changes in the number of different types of nodes during the process of forming a star cluster structure with the iteration process, as well as the number of times the four designed coalition operations are selected in each iteration.

[0185] At the initial moment of simulation, each satellite node is set as an isolated node and a cluster head node. At this time, there are no satellites serving as member nodes. Therefore, at the initial stage of the simulation, the number of isolated nodes is equal to the number of cluster head nodes, while the number of member nodes is zero. As the clustering algorithm continues to run, satellite nodes gradually join and form clusters, resulting in a rapid decrease in the number of isolated nodes and cluster head nodes. At the same time, the number of member nodes continues to increase. When the simulation time reaches approximately 40 seconds, the number of various types of nodes tends to stabilize, indicating that the algorithm finally forms a Nash-stable cluster structure. In this stable cluster structure, the final number of cluster head nodes is 20, and the number of isolated nodes drops to zero, which means that 336 satellite nodes are effectively divided into 20 clusters. At this time, there are no nodes forming separate clusters in the cluster structure, thus verifying that the clustering algorithm has good convergence performance and clustering effect in this scenario.

[0186] Figure 6 (b) shows the number of times the designed coalition operations are selected over time during the operation of the self-organization algorithm to analyze the applicable clusters of different operations and the rationality of the operation design. According to the CSOC-G algorithm setting, an operation will only be selected when the operation gain that the node can obtain under the current time for this operation is the highest. It is not difficult to see from the figure that all operations have been selected during the algorithm operation, reflecting the rationality of the design of each operation. The analysis of each operation is as follows:

[0187] Coalition conversion operation: This operation involves nodes joining other clusters to achieve the rapid merger of satellite nodes. In the initial stage of the algorithm execution, the coalition conversion operation is frequently selected because in the initial stage of cluster formation, the need for node merger is relatively urgent.

[0188] Bilateral exchange operation: As the algorithm runs to the middle and late stages, the cluster structure tends to be stable, and the selection frequency of the bilateral exchange operation begins to increase. This may be because the number of nodes in some clusters has reached the upper limit, and new nodes cannot join these clusters through the coalition conversion operation (even if joining can significantly improve the objective function). Therefore, through the bilateral exchange operation, the nodes in the clusters need to be exchanged to achieve the addition of these clusters without violating the cluster node number limit.

[0189] Cluster head election operation: In the initial stage of iteration, due to the large changes in the cluster structure, changing the cluster head node can bring higher benefits, so the cluster head election operation is frequently selected. However, as the iteration progresses, the overall change in the cluster structure decreases, and the number of times the cluster head election operation is selected also decreases.

[0190] Improved replacement operation: This operation is designed specifically for isolated nodes and is selected relatively few times during the entire algorithm operation. In the middle stage of the algorithm iteration, such as Figure 6(As shown in (a)), there are still a certain number of isolated nodes in the network. These nodes can improve the coalition value by replacing nodes in the star clusters whose quantity has reached the upper limit. In this case, the coalition conversion operation is no longer applicable, and the replacement operation needs to be improved to replace the nodes in the saturated clusters with isolated nodes.

[0191] Through the above analysis, this study not only reveals the roles and applicability of different operations in the process of star cluster structure formation, but also verifies the rationality and effectiveness of the designed operations.

[0192] Next, a set of comparative experiments are used to illustrate the effectiveness of the self-organizing clustering algorithm proposed in the present invention. The comparative algorithms include: Coalition Selection Algorithm based on Coalition Order (CSA-CO), Reliable and Low-overhead Clustering Scheme (RLOC), Coalition Selection Algorithm based on Coalition Expected Altruistic Order (CSA-CEAO), and Merge / Split / Transfer Coalition Formation Algorithm (M / S / T-CFA). Among them, the first three are star cluster self-organizing algorithms, and the last one is a centralized solution algorithm. The detailed descriptions of these comparative algorithms are as follows:

[0193] CSA-CO: This algorithm makes selections based on the coalition order. First, bilateral exchange operations are performed, and then coalition exchange operations are executed to form a set of feasible operations. The coalition preference order is used as the preference order of the participants, and then a key operation is randomly selected from this set, and the next operation is updated according to the log-linear learning strategy.

[0194] RLOC: Only coalition exchange operations are included in its coalition formation rule. In each round of iteration, all feasible coalition exchange operations of the search nodes are searched, and the operation with the largest operation benefit is selected to be accepted and the next operation is updated.

[0195] CSA-CEAO: Similar to CSA-CO, only coalition exchange operations are included in its coalition formation rule. A coalition preference order based on the coalition expected altruistic order is designed, and based on this, an operation is randomly selected from the set of feasible operations for acceptance and update.

[0196] M / S / T-CFA: In the alliance formation rules, there are alliance merger operations, alliance split operations, and alliance transformation operations. The alliance structure is updated in the order of alliance merger / split / transformation. However, since this update is centralized, it is difficult to define a convergence curve. Therefore, in this study, only the final results are compared, and the comparison of the convergence curve is not involved.

[0197] In the solution of the present invention, according to different operation selection strategies, the proposed CSOC algorithm is divided into two variants. Among them, the algorithm that uses the greedy strategy for alliance operation selection is called CSOC-G, and the clustering algorithm that uses the log-linear learning strategy (LLA) for alliance operation selection is denoted as the CSOC-LLA algorithm. The value of the learning parameter ε is set to 0.85.

[0198] Generally speaking, when there is an initial alliance structure in the scenario, the objective values that the algorithm can achieve are generally lower than those in the scenario without an initial cluster structure. The reason for this phenomenon may be that the initial cluster structure itself is not the optimal solution, and its existence may affect the further optimization of the subsequent cluster structure. The CSOC-G algorithm and the CSOC-LLA algorithm have achieved significantly better objective values than other algorithms in all test scenarios. In particular, in Scenario 4 with a large number of nodes, the CSOC-LLA algorithm can obtain a slightly higher objective value than the CSOC-G algorithm. The performance of the M / S / T-CFA algorithm is second only to that of the CSOC-G algorithm and the CSOC-LLA algorithm in most scenarios.

[0199] Above, for the self-organizing clustering problem of large-scale constellation networks, the self-organizing clustering algorithm based on alliance formation game provided by the present invention has the best effect and obvious advantages in reducing the cluster communication overhead and improving the cluster cooperation ability, etc., and has important value for further exerting the advantages of inter-satellite links and improving the cooperation efficiency of large-scale constellation systems.

[0200] Those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computer device. Optionally, they can be implemented by program codes executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be separately made into individual integrated circuit modules, or multiple modules or steps among them can be made into a single integrated circuit module to implement. The present invention is not limited to any specific combination of hardware and software.

[0201] Although the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, they are not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or deformations that can be made without creative efforts on the basis of the technical solutions of the present invention are still within the scope of protection of the present invention.

[0202] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or equivalently replace some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A large-scale star cluster network self-organizing clustering method, characterized in that: Follow these steps: S1. Obtain the position data of all satellites in the ground station server and identify different satellites in the same layer; S2. Establish a mathematical model of self-organizing clustering for large-scale star clusters; S21. Perform constellation network system modeling; S22. Determine the self-organizing clustering benefit function for large-scale star clusters; S23. Form a mathematical model of self-organizing clustering for large-scale star clusters; S3. Construction method of large-scale star clustering model based on alliance formation game; S31. Game model based on alliance formation; S32. Determine the rules for forming an alliance; S33. Determine the preference order of participants; S4. Self-organizing clustering algorithm based on alliance formation game; S41. Build an asynchronous decision-making algorithm framework; S42. Forming a self-organizing clustering algorithm A1; S43. Forming a self-organizing clustering algorithm A2; According to whether the cluster to which node i currently belongs satisfies the CFPLS problem constraints, the designed clustering algorithm A1 and clustering algorithm A2 are executed respectively; If the cluster to which node i currently belongs satisfies the problem constraints in CFPLS, that is, if and only if the cluster C to which node i belongs m satisfy When , node i executes clustering algorithm A1, otherwise it executes clustering algorithm A2; Among them, in the problem constraints, N max and D max are the maximum limits for cluster size and cluster diameter, respectively.

2. A large-scale star cluster network self-organizing clustering method as claimed in claim 1, characterized in that: In step S21, Different communication links use different communication methods. There are differences in the transmission rates of intersatellite transmissions.

3. A large-scale star cluster network self-organizing clustering method as claimed in claim 2, characterized in that: In step S21, The orbital radius of satellites i and j in the same layer is the same, that is, Link transmit power P t , maximum directional antenna gain G0 and channel power gain h i,j , the same-layer intersatellite link transmission rate rt i,j for: Where K and T represent the Boltzmann constant and the total system noise temperature, respectively. It represents the ratio requirement of received energy per bit to noise power, and M represents the link margin.

4. A large-scale star cluster network self-organizing clustering method as claimed in claim 1, characterized in that: In step S21, L(d i,j ) represents the distance-dependent free-space propagation loss, where d i，j is the distance between satellite i and satellite j. Given the speed of light c and the frequency of the communication electromagnetic wave F, the calculation formula is as follows: The orbital radii of satellites i and j in different layers are different, that is, 5. A large-scale star cluster network self-organizing clustering method as claimed in claim 4, characterized in that: In step S21, The channel bandwidth B of the intersatellite link can be calculated using the Shannon formula to achieve the transmission rate rt i,j , where rt i,j is a parameter: rt i,j =B log(1+Λ i,j ) Among them, Λ i,j represents the signal-to-noise ratio (SINR) of the intersatellite link between satellite i and satellite j.

6. A large-scale star cluster network self-organizing clustering method as claimed in claim 5, characterized in that: In step S21, The intersatellite link transmission rate rt between any two satellites i and j i,j It is expressed as:

7. A large-scale star cluster network self-organizing clustering method as claimed in claim 6, characterized in that: In step S21, When an available intersatellite link cannot be directly established between satellite i and satellite j, a multi-hop intersatellite link is established through other satellite nodes as relays; the link capacity of the multi-hop intersatellite link is calculated based on all the intersatellite links passed; The shortest path between satellite i and satellite j is denoted as q i,j =((i,i1),(i1,i2),...,(i h ,j)), then the link capacity of the multi-hop intersatellite link between satellite i and satellite j is: Among them, α is the loss coefficient of the multi-hop path, usually α < 1. The more the path jumps, the greater the link capacity loss. i,j |=1, which means that the link between nodes does not need to be transferred. At this time, κ i,j =rt i,j , κ i,j It represents the link capacity of the inter-satellite link between any two satellites in the constellation network.

8. A large-scale star cluster network self-organizing clustering method as claimed in claim 7, characterized in that: In step S21, The constellation network is represented as an undirected graph (V, E), where V represents the set of satellite nodes and E represents the set of edges between satellites. i,j = 1, (i, j)∈E, satellite i and satellite j can be regarded as neighbor satellites, and the inter-satellite link relationship of a large-scale constellation network can be expressed by the adjacency matrix l:

9. A large-scale star cluster network self-organizing clustering method as claimed in claim 8, characterized in that: In step S21, For each edge in E, the capacity value of the corresponding intersatellite link can be calculated; The inter-satellite link capability of a large-scale constellation network is represented by the link capability matrix κ:

10. A large-scale star cluster network self-organizing clustering method as claimed in claim 9, characterized in that: In step S22, In the cluster structure Π of a large-scale star cluster network, the collaborative ability χ total It is expressed as: Among them, χ inter (C m ) and χ intra (C m ) represent clusters C m The inter-cluster coordination ability and intra-cluster coordination ability of inter (C m ) is related to the link capacity of the inter-satellite link between cluster head nodes, j∈(V Head ∩C m ) represents cluster C m The cluster head node, V Head represents the set of cluster head nodes in the cluster structure, β inter is a parameter to adjust the importance of inter-cluster synergy, indicating the proportion of inter-cluster synergy to the total cluster structure synergy; intra-cluster synergy χ intra (C m ) is related to the link capability between members within the cluster. When the intersatellite links between members within the cluster have stronger link capabilities, the intra-cluster coordination capability is also stronger.

Images