Dynamic management topology construction, evolution, and maintenance of low Earth orbit mega-constellation

Low earth orbit (LEO) mega-constellations, characterized by dense satellite coverage and low transmission latency, have substantial advantages in global communications, earth observation, and disaster forecasting. Compared to traditional constellations with simple configurations, mega-constellations have a large number of nodes and complex satellite interaction relationships and are oriented to vast global missions. As a result, when the traditional ground-based centralized management mode is applied to a mega-constellation, the ground station suffers from an explosive growth of control load and struggles to achieve the efficient, timely, and orderly operation of the mega-constellation. Therefore, it is necessary to adjust the management architecture to partially transfer the on-ground management mission to on-board management mission. Clustering satellites into multiple management domains and implementing unified management within each domain is an effective method for realizing on-board management of mega-constellations. Existing studies on the construction of satellite management domains have focused on some specific indicators, such as spatial distribution uniformity, network control latency, and controller load. Considering the highly dynamic motion of the mega-constellation, a fixed management domain construction will inevitably lead to frequent updates of the management architecture, thereby increasing the burden of on-board management. Thus, more stable strategies for constructing and updating the management domain need to be explored to reduce the frequency of management architecture updates over extended mission periods. In a research article recently published in Space: Science & Technology, scholars from Nanjing University of Aeronautics and Astronautics, Harbin Institute of Technology (Shenzhen), and Politecnico di Milano together proposed a management strategy using distributed management domains as well as their dynamic evolution and maintenance methodology which ensures low transmission latency and management update frequency and facilitates intra-domain autonomous management and inter-domain information interaction, thereby laying the foundation for long-term autonomous management of the mega-constellation.

First, the topological definition of the management domain, the management topology construction problem model focusing on minimizing transmission latency and update frequency, and performance evaluation of management topology are presented. The management domain model based on the software-defined networking (SDN) is shown in Fig. 1. In the control plane, the center node handles network state monitoring, routing calculation, resource scheduling, command distribution, and other functions related to autonomous network operation and maintenance, thus partially replacing the current function of ground station. The center node manages member nodes via one-hop control links and communicates with other center nodes within communication range. In the data plane, member nodes are linked by data links. The above 2 planes do not overlap with each other. The management domains construction mainly focuses on the control plane.

The Walker constellation comprising P planes with N satellites in each plane is considered, in which all satellites are assumed to have the same on-board resource. The set S_whole = {s₁, s₂, ⋯ s_i, ⋯ s_m} denotes all satellites in the mega-constellation. The distance between each satellite pair as a metric is attached to the set S_whole. S^c = {s_i^c | s_i^c ∈ S_i, 1 ≤ i ≤ n} denotes a total of n center nodes arranged across the constellation. S_i = {s_i^c, s_i¹, s_i², ⋯, s_i^j} (i = 1, 2, ⋯, n) denotes all m satellites divided into n management domains without overlapping members. Thus, the management topology is S_manage = {S₁, S₂, ⋯, S_n} where ∀ S_i, S_j ∈ S_manage, S_i ∩ S_j = ∅. The management relationship between the center node and member satellites in S_i is represented by the graph G_i = (S_i, E_i, t), t₀ ≤ t ≤ t₀ + T_m, where t₀ is the initial time of management and T_m is the management period. E_i = { (s_i^c, s_i^j) | s_i^c ∈ S^c, s_i^j ∈ S_i\s_i^c} is the set of control links connecting the center node to its associated members. Therefore, the management architecture construction and maintenance can be described as the following problem: dividing all satellites in S_whole into multiple time-varying nonoverlapping subsets S_i by determining the center node set S^c, as well as the management relationship G_i to form the time-varying management topology S_manage according to the satellite motion.

Furthermore, considering the connectivity constraints, the dynamic property of the orbital motion, the time of management switching of member satellite, the management completion constraint for management topology construction, the management topology construction and maintenance problem can be modeled as Eq. (17):

The performance of management topology is evaluated by the global average transmission latency L(t), the frequency of satellite management affiliation switching K_c(t), and the member quantity uniformity K_q.

Then, the management topology construction and maintenance strategy including center node arrangement, optimal member assignment, and dynamic management affiliation switching are investigated. Center nodes are arranged evenly throughout the constellation to minimize the average latency in intra-domain transmission, which involves 2 steps: the parallel derivation and the same-orbit derivation, as shown in Fig. 2. The parallel derivation generates a ring of center nodes that are uniformly distributed and parallel to the latitude line. The number of center nodes in parallel derivation I is constrained as I = 360°/α, where α is the central angle corresponding to the two neighboring center nodes. For the same orbit derivation, every center node in the parallel derivation is chosen as the first node in its orbit, and subsequent center nodes in each orbit will be generated similarly with a discrepancy angle β. The number of nodes J in each orbit is J = 360°/β. Then, n = I × J center node set S^c is generated. The complete steps of the management topology construction for the entire constellation are illustrated in Fig. 3. The management affiliation for any s_j in S_whole\S^c is assign based on minimizing F(s_i^c, s_i^j) (in Eq. (17)) after determining the center node set S^c. The process of autonomous management maintenance for satellite s_j is shown in Fig. 4. If satellite s_j detaches from the original management domain before the end of management update period T_m, it needs to be reclassified into a new management domain. After the reclassification, s_j predicts the t_detach relative to the new center node. If t_detach < t₀ + T_m, a new round of management maintenance is turned to in the detachment moment.

Last, simulation results and the conclusion are presented. Based on the Starlink constellation, a total of 1248 operational satellites in the constellation with fixed constellation configuration are screened for simulation. Fig. 5 shows the distribution of management domains with center nodes on the same orbit. Fig. 6 illustrates the spatial distribution of management domains. Results show that all satellites are organized into 81 management domains. The management span of topology K_g = 100%. The average intra-domain transmission latency L(t₀) = 4.8 ms. The standard deviation the number of members in each domain is K_q = 3.08, where the average is 15. The management topology construction ensures management load balancing while achieving low latency for intra-domain management. Further, the management period is set to T_m = 1 month, and the constellation motion is simulated using the J2 perturbation orbital propagation model. Fig. 7 illustrates the effect of management maintenance on the average transmission latency of the mega-constellation. Without management maintenance, latency increases rapidly from 4.7 ms initially to 18.4 ms within a month. With management maintenance, the average transmission latency is consistently maintained within 4.7 to 7.8 ms, demonstrating the effectiveness of management maintenance. In addition, a connection between the intrinsic properties of the constellation and the management architecture can be found: 1. Mega-constellations are primarily designed using walker and polar configurations, it is easy to achieve a uniform management domain distribution for such mega-constellations with a dense distribution of satellites. 2. The cyclical operation of mega-constellation makes dynamic updates to the management architecture predictable. 3. The management architecture facilitates autonomous information interactions on board, assuming that the mega-constellation is equipped with inter-satellite links.