A cooperative hop-beam method suitable for large-scale multi-layer low earth orbit satellite constellation

Abstract

The application discloses a kind of cooperation jump beam methods suitable for large-scale multilayer low-orbit satellite constellation, belong to low-orbit satellite field;Specifically, first, the earth surface is evenly divided, and the communication resource request in each region is divided by threshold;Then, according to the mapping relationship of satellite foot point, the satellite set of each partition is obtained, as initial star cluster;And calculate the ratio Δ of the number of satellites in each partition and the service density, according to Δ, borrow the edge satellite of adjacent area, until reaching load balancing, form the final service star cluster;Then, according to the service classification of cell, respectively select the service satellite of each cell.Last, using deep DQN network learning is carried out to the satellite in star cluster Jump beam resource collaborative scheduling, finally selected to meet the different user QoS demand and avoid inter-beam interference Jump beam scheme.The application alleviates the problem of large difference between beam interference and traffic volume faced by multi-satellite jump beam.

Description

Technical Field

[0001] This invention belongs to the field of low-Earth orbit satellites, and specifically relates to a cooperative beam-hopping method applicable to large-scale multi-layer low-Earth orbit satellite constellations. Background Technology

[0002] Low-Earth orbit (LEO) satellites offer advantages such as low latency, low cost, and flexible networking, making them promising for widespread applications. However, due to the limited onboard resources of a single LEO satellite and the rapid movement of its service area, uninterrupted communication in a specific area on the ground is difficult to achieve, thus failing to adequately meet the QoS requirements of ground users. Therefore, current LEO satellite communication systems typically employ LEO satellite constellations to increase ground coverage, expand system capacity, and collaboratively complete communication tasks.

[0003] Low Earth Orbit (LEO) satellite constellations consist of a certain number of LEO satellites and have a relatively stable spatial geometry. Currently, the development trend of LEO satellite constellations is from single-layer to multi-layer and from small-scale to large-scale. For example, the first-generation Iridium constellation used only 66 satellites to achieve Earth coverage, while the second-generation OneWeb constellation uses 720 LEO satellites, and Starlink's satellite count ranges from hundreds to thousands. As the deployment scale of LEO satellite constellations continues to expand and service demands become more diversified and complex, dividing the constellation into several clusters of satellites that are relatively close in space and have collaborative functional relationships can reduce the complexity and cost of satellite network management and improve the stability of constellation operation and maintenance.

[0004] Traditional medium- and high-orbit satellite-to-ground communication typically employs a single beam or fixed multiple beams, with fixed resource allocation, which can easily lead to resource redundancy. Beam hopping technology can dynamically allocate onboard communication and bandwidth resources, and rationally schedule beams in a time-division multiplexing manner to achieve the goal of serving multiple cells with a small number of beams. It has high timeliness and flexibility and has attracted widespread attention from scholars at home and abroad.

[0005] However, existing low-Earth orbit (LEO) satellite resource allocation methods do not consider load balancing among satellites. Due to the time-varying nature of ground user services and their uneven geographical distribution, densely populated service areas face service delays while sparsely populated service areas experience excessive resource redundancy. Furthermore, most existing beam-hopping design methods only consider beam hopping within a single satellite. For LEO satellite constellations, the beam resource allocation of different satellites affects each other, and the interference between different satellite beams cannot be ignored. Therefore, the overall constellation optimization objective urgently needs to be considered.

[0006] Therefore, researching a multi-satellite cooperative beam-hopping method based on regionally associated satellite clustering can achieve inter-satellite load balancing while flexibly and quickly scheduling beam resources, isolating interference, and meeting the QoS requirements of different users, which is of great significance to the development of satellite communication.

[0007] Due to the uneven spatial distribution of services provided by low-Earth orbit (LEO) satellites and the dynamic, time-varying nature of satellite communication resources, there exist areas with dense and sparse services. For multi-tiered LEO satellite constellations, there are often areas where multiple satellites overlap in coverage. If users in adjacent hotspot areas are connected to the nearest satellite, that satellite will be fully loaded or overloaded; conversely, for satellites with sparse services within their coverage areas, their onboard resources cannot be utilized efficiently. Therefore, simply mapping ground cells to satellites based on distance is not the optimal allocation solution. Satellites must comprehensively consider the current status of their onboard resources when determining their service cells to achieve load balancing.

[0008] Furthermore, with the continuous expansion of satellite deployment, traditional planar management structures suffer from high complexity, high latency, and unbalanced load. Effective satellite clustering strategies offer the possibility of reducing satellite network management complexity and balancing inter-satellite load. However, current clustering strategies mostly focus on network topology and do not fully consider the differences in services across geographical regions in real-world application scenarios, requiring further optimization. Achieving a balance between the number of satellites in a constellation and service requests from associated regions through low-Earth orbit satellite clustering, and providing more satellite resources to hotspot areas, is key to solving the load balancing problem.

[0009] Meanwhile, most existing beam-hopping algorithms are designed for single satellites and lack in-depth consideration of beam resource scheduling schemes in multi-satellite collaborative scenarios. Compared with traditional single-satellite beam-hopping strategies, beam resource scheduling for multiple satellites usually requires consideration of factors such as inter-satellite beam interference and load balancing, and requires collaborative design of beam patterns through inter-satellite information exchange.

[0010] Traditional satellite beam allocation methods, such as polling scheduling and fixed time slot beam allocation, can solve the problem of low flexibility in fixed beam allocation to some extent with lower computational complexity by using time division multiplexing technology. However, they are still difficult to meet the QoS requirements of users in future communication environments.

[0011] At present, artificial intelligence technologies, represented by deep reinforcement learning, have been widely used in the field of satellite communication. Using deep networks with characteristics such as intelligence, self-learning, and high dynamics to solve beam allocation problems can achieve superior system performance with a shorter computation time and better coordinate and manage the limited beam resources of satellite constellations. Summary of the Invention

[0012] To address the aforementioned issues, this invention proposes a cooperative beam-hopping method suitable for large-scale, multi-layered LEO satellite constellations. Using a large-scale, multi-layered LEO satellite constellation as the application scenario, the ground is dynamically divided into zones of varying sizes based on regional traffic volume, and satellite clusters are also created. Each zone is provided with communication access by a satellite cluster, and satellites within a cluster perform intelligent beam scheduling through multi-satellite cooperation. The formation of satellite clusters is directly related to ground service requests. For different service areas, satellites can dynamically adjust their beamwidth based on the volume of service requests. By jointly considering satellite clustering algorithms and beam-hopping algorithms, beam allocation and scheduling are performed with the goal of fully utilizing onboard resources and meeting user QoS requirements.

[0013] The cooperative beam-hopping method applicable to large-scale multi-layer low-Earth orbit satellite constellations comprises the following specific steps:

[0014] Step 1: Divide the Earth's surface evenly into areas S, and obtain the communication resource request size γ in each area, using a threshold γ... threshold As a partitioning standard.

[0015] The communication resource request γ≥γ threshold The areas designated as business-intensive areas are further divided into separate zones.

[0016] The communication resource request γ < γ threshold Regions with sparse business operations are called business-sparse regions. These regions are selected from neighboring regions of the same type and merged to form new partitions. The specific number of neighboring regions of the same type to be merged is determined by a threshold γ. threshold The decision is made when the total communication resource requests of the sparse service area and all neighboring areas of the same type exceed the threshold γ. threshold This means the merger is complete.

[0017] After several iterations, the ground communication area was divided into zones of varying sizes.

[0018] Step 2: Obtain the satellite set within each partition based on the mapping relationship of the satellite sub-satellite points, which serves as the initial satellite cluster corresponding to that partition.

[0019] For the m-th partition, if it is a business-intensive partition, its area is the same as that of the basic partition, i.e., S. m =S, Number of serving satellites N m =N s ;

[0020] If it is a sparse business partition formed by merging k basic partitions, then the area S m =kS, the number of serving satellites satisfies N m =kN s .

[0021] Where, N sN represents the number of serving satellites corresponding to an area of ​​size S; m This indicates the number of service satellites corresponding to partition m.

[0022] Step 3: Calculate the ratio Δ of the number of satellites to the service density in each partition corresponding to each initial cluster, and sort the service-intensive partitions in ascending order according to Δ; for the service-intensive partitions, borrow edge satellites from adjacent areas in turn until the load of all partitions is balanced to form the final service cluster;

[0023] The ratio Δ of the number of satellites to the service density is calculated using the following formula:

[0024]

[0025] Where D represents the business density of the region.

[0026] The satellite secondment process is as follows:

[0027] First, for the service-intensive partition with the smallest current satellite count to service density ratio Δ Select the service-sparse partition with the largest Δ among its adjacent partitions. And divided into business-sparse partitions Towards business-intensive partitions Provide satellites;

[0028] Then, after each satellite is borrowed, update the number of satellites and the Δ table for both the current service-intensive and service-sparse partitions. Select the service-intensive partition with the smallest updated Δ. Borrowing the business sparse partition with the largest Δ among adjacent partitions The satellites in the system will be seconded, and the Δ table will be updated again.

[0029] Until the regional load balancing of all partitions is satisfied The final service cluster can then be obtained;

[0030] This represents the threshold.

[0031] Step 4: For each service cluster, select the appropriate service satellites for each cell within the cluster based on the service classification of the cells within its coverage area.

[0032] The intra-cluster cell selection process is as follows:

[0033] First, regarding the current community C n The visible satellite set V corresponding to the cell is determined based on the geographical relationship between the satellite and the cell. n ;

[0034] Then, determine cell c. nWhether a cell belongs to the edge or center of a zone is determined by the edge cell's preference for the serving satellite.

[0035] For edge cells:

[0036] If the community belongs to a service-intensive zone, the number of visible satellites is V. i Compared to before the satellite borrowing, there is an increase in the number of low-level low-orbit satellites borrowed or those from the original cluster. By selecting the satellite with the least load, load balancing and latency requirements of each satellite are achieved.

[0037] If a cell belongs to a service-sparse zone and satellites within that zone are reassigned, then in cells without low-Earth orbit (LEO) satellite coverage, visible high-Earth orbit (LEO) satellites will be selected to supplement coverage, meaning that high-Earth orbit (LEO) satellites will provide services to that cell.

[0038] For the central residential area:

[0039] If the cell belongs to a service-sparse zone, the original visible low-level low-orbit satellite will be selected based on load balancing and service classification within the cell.

[0040] If the community belongs to a service-intensive zone, load balancing strategies can be used during satellite selection to select existing low-level low-Earth orbit satellites or high-level low-Earth orbit satellites as supplementary coverage.

[0041] This process continues until all communities have been visited and the serving satellites for each community have been identified.

[0042] Step 5: Treat each satellite in the cluster as an intelligent agent, responsible for beam scheduling of the corresponding serving cell. Use a deep DQN network to learn and coordinate the beam hopping resources of the satellites in the cluster, and finally select a beam hopping scheme that meets the QoS requirements of different users and avoids inter-beam interference.

[0043] In the deep DQN network learning process

[0044] (1) The state is defined as follows:

[0045]

[0046] in, This indicates the state of the l-th satellite within the cluster. This represents the cluster state information collected by the star cluster. Indicates real-time business. This indicates non-real-time business.

[0047] (2) The action is defined as follows:

[0048]

[0049] in, N represents the beam scheduling status of the l-th satellite. S Indicates the total number of satellites; φ n =1 indicates that the nth cell was selected; N C This indicates the total number of cells; K represents the number of beams that the satellite can provide.

[0050] (3) Reward: Three metrics are used to provide reward value for the agent.

[0051] First, the latency reward for real-time data packets is defined as follows:

[0052] in Indicates the latency of data packets. This represents the total latency of all real-time data packets within the coverage area of ​​a single satellite.

[0053] Secondly, using time slot t i The total amount of non-real-time data packets leaked is used as its throughput reward, defined as follows:

[0054]

[0055] Indicates time slot t i satellite beam pattern, Indicates time slot t i Channel capacity of the beam service cell.

[0056] Finally, the spatial isolation value, shared by all satellites within the cluster, is defined as the interference avoidance reward as follows:

[0057]

[0058] H represents the adjacency matrix of each cell. x i ∈{0,1} represents the binary vector of the hopping beam pattern of a single satellite, x i =1 indicates cell c i Subject to beam service. When the service beams of a single satellite all meet the space isolation condition, there should be x. T Hx = 0.

[0059] The advantages of this invention are:

[0060] 1) A cooperative beam-hopping method applicable to large-scale multi-layer low-Earth orbit satellite constellations. Based on the differences in ground service density, the large-scale low-Earth orbit satellite constellation is divided into several satellite clusters. The on-board load is balanced through edge satellite borrowing strategy and satellite cell correspondence strategy, so as to achieve load balancing between and within clusters and alleviate the problems of inter-beam interference and large differences in service volume faced by multi-satellite beam-hopping.

[0061] 2) A cooperative beam-hopping method applicable to large-scale multi-layer low-Earth orbit satellite constellations. It uses a deep DQN network with self-learning and high dynamics to solve the beam-hopping design problem. Compared with traditional heuristic algorithms, it can achieve better beam-hopping system performance with shorter computation time.

[0062] 3) A cooperative beam-hopping method applicable to large-scale multi-layer low-Earth orbit satellite constellations is proposed. A novel intelligent beam-hopping scheme based on regionally associated satellite clustering is proposed, providing an effective solution to the problem of multi-satellite resource allocation for differentiated services for ground users. Attached Figure Description

[0063] Figure 1 This invention presents a multi-satellite dynamic cooperative beam-hopping system model based on a regionally associated clustering strategy.

[0064] Figure 2 This is a flowchart of a cooperative beam-hopping method applicable to large-scale multi-layer low-Earth orbit satellite constellations according to the present invention;

[0065] Figure 3 This is a flowchart of the edge satellite borrowing method of the present invention;

[0066] Figure 4 This is a schematic diagram of the regional clustering formed by borrowing edge satellites in this invention;

[0067] Figure 5 This invention relates to a multi-agent deep learning architecture; Detailed Implementation

[0068] The embodiments of the present invention will now be described in detail and clearly with reference to the accompanying drawings.

[0069] This invention proposes a cooperative beam-hopping method suitable for large-scale multi-layered low-Earth orbit satellite constellations. It employs a multi-satellite cooperative beam-hopping algorithm with large-scale multi-layered low-Earth orbit satellite constellations as the application scenario. Figure 1 As shown. To provide more targeted services to users, this invention dynamically divides the ground into zones of varying sizes based on the volume of regional traffic, and simultaneously divides them into satellite clusters. Each zone is provided with communication access by a satellite cluster, and satellites within a cluster perform intelligent beam scheduling through multi-satellite cooperation. The formation of satellite clusters is directly related to ground service requests, and for different service areas, satellites can dynamically adjust their beamwidth according to the volume of service requests.

[0070] The cooperative beam-hopping method applicable to large-scale multi-layer low-Earth orbit satellite constellations, such as Figure 2 As shown, the specific steps are as follows:

[0071] Step 1: Divide the Earth's surface evenly into areas S, and obtain the communication resource request size γ in each area, using a threshold γ... threshold As a partitioning standard.

[0072] The communication resource request γ≥γ threshold The areas designated as business-intensive areas are further divided into separate zones.

[0073] The communication resource request γ < γ threshold Regions with sparse business operations are called business-sparse regions. These regions are selected from neighboring regions of the same type and merged to form new partitions. The specific number of neighboring regions of the same type to be merged is determined by a threshold γ. threshold The decision is made when the total communication resource requests of the sparse service area and all neighboring areas of the same type exceed the threshold γ. threshold This means the merger is complete.

[0074] After several iterations, the ground communication area was divided into zones of varying sizes.

[0075] Step 2: Obtain the set of satellites in each partition region based on the mapping relationship of satellite sub-satellite points, which serves as the initial satellite cluster corresponding to that partition.

[0076] Assuming the satellites are relatively uniformly distributed over the ground, meaning the number of serving satellites is equal within each basic partition of area S. For the m-th partition, if it is a service-intensive partition, then its area is the same as the basic partition, i.e., S. m =kS, number of service satellites N m =N s ;

[0077] If it is a sparse business partition formed by merging k basic partitions, then the area S m =kS, the number of serving satellites satisfies N m =kN s .

[0078] Where, N s N represents the number of serving satellites corresponding to an area of ​​size S; m This indicates the number of service satellites corresponding to partition m.

[0079] Step 3: Calculate the ratio Δ of the number of satellites to the service density in each partition corresponding to each initial cluster, and sort the service-intensive partitions in ascending order according to Δ; for service-intensive partitions, borrow edge satellites from adjacent areas in turn until the load of all partitions is balanced, forming the final service cluster of that geographical partition;

[0080] The edge satellite borrowing strategy is key to the regional clustering scheme; the ratio Δ of the number of satellites to the service density of a certain partition is defined and calculated as follows:

[0081]

[0082] Where D represents the service density of the region. A larger Δ indicates that satellite resources within that region are more likely to become idle, allowing satellites to be loaned to other regions.

[0083] like Figure 3 As shown, the satellite borrowing process is as follows:

[0084] First, for the densely populated partition with the smallest ratio Δ between the current number of satellites and the service density. Select the service-sparse partition with the largest Δ among its adjacent partitions. And divided into business-sparse partitions Towards business-intensive partitions Provide satellites;

[0085] Then, after each satellite is borrowed, update the number of satellites and the Δ table for the current service-intensive and service-sparse partitions. Select the service-intensive partition with the lowest updated density Δ. The sparse region with the largest ratio Δ of borrowing from adjacent sparse partitions. The satellites in the system will be seconded, and the Δ table will be updated again.

[0086] Until the regional load balancing of all partitions is satisfied This represents the threshold. The final service clusters provided by this partition can then be obtained; the final service cluster result is as follows: Figure 4 As shown.

[0087] Step 4: For each service cluster, select the appropriate service satellites for each cell based on the service classification of the cells within its coverage area.

[0088] The intra-cluster cell selection process is as follows:

[0089] First, regarding the current community C n The visible satellite set V corresponding to the cell is determined based on the geographical relationship between the satellite and the cell. n ;

[0090] Then, determine cell c. n Belongs to partition Ω m Whether a cell is located on the periphery or in the center, the peripheral cell will have priority in selecting the serving satellite.

[0091] For edge cells:

[0092] If the community belongs to a service-intensive zone, the number of visible satellites is V. i If the number of satellites has increased compared to before the borrowing, then, taking into account indicators such as load balancing and time delay requirements, priority should be given to borrowed low-level low-orbit satellites or low-level low-orbit satellites within the original cluster.

[0093] Prioritize selecting satellites with the lowest existing load, that is, to achieve load balancing for each satellite as much as possible.

[0094] If a cell belongs to a service-sparse zone and satellites within that zone are reassigned, then in cells without low-Earth orbit (LEO) satellite coverage, visible high-Earth orbit (LEO) satellites will be selected to supplement coverage, meaning that high-Earth orbit (LEO) satellites will provide services to that cell.

[0095] For the central residential area:

[0096] If the cell belongs to a service-sparse zone, the original visible low-level low-orbit satellite will be selected based on load balancing and service classification within the cell.

[0097] If the community belongs to a service-intensive zone, load balancing strategies can be used during satellite selection to select existing low-level low-Earth orbit satellites or high-level low-Earth orbit satellites as supplementary coverage.

[0098] This process continues until all communities have been visited and the serving satellites for each community have been identified.

[0099]

[0100] Step 5: Treat each satellite in the cluster as an intelligent agent, responsible for beam scheduling of the corresponding serving cell. Use a deep DQN network to learn and coordinate the beam hopping resources of the satellites in the cluster, and finally select a beam hopping scheme that meets the QoS requirements of different users and avoids inter-beam interference.

[0101] Due to the uneven distribution of terrestrial users and the diverse needs of users, different service types have different requirements for QoS parameters such as transmission latency and throughput. Therefore, when allocating beams for satellite access users, it is necessary to consider the needs of each user. For latency-sensitive user services, priority should be given to serving them, while for non-latency-sensitive services, the communication throughput should be maximized. At the same time, satellite onboard resources are limited. When scheduling beams and allocating resources, the load capacity of each satellite and the interference between beams must be considered, ensuring that the onboard resources of each satellite are utilized evenly within their load capacity to reduce interference.

[0102] The beam-hopping scheme of this invention will jointly consider satellite clustering algorithm and beam-hopping algorithm to make full use of onboard resources and meet user QoS requirements for beam allocation and scheduling: each satellite in the cluster is regarded as an intelligent agent, responsible for the beam scheduling of the corresponding serving cell. While each intelligent agent makes independent actions, it can share the global state and reward. The state and actions are sent to the Q network for training, and finally, a beam-hopping scheme that meets the QoS requirements of different users and avoids inter-beam interference can be selected.

[0103] like Figure 5As shown, the specific steps of the DQN-based hopping beam pattern design scheme are as follows:

[0104] Define the amount of data stored in each cell in the on-board memory. The number of services arriving in each time slot is expressed as follows: The amount of data stored in the satellite's memory can then be represented as:

[0105]

[0106] in Indicates time slot t i satellite beam pattern, Indicates time slot t i Channel capacity of the beam-serving cell.

[0107] Define state s i ∈S, action selection a i ∈A and reward R(s) i ,a i The specific definitions of each element are as follows:

[0108] (4) State: If a satellite cluster is viewed as an environment, then the environment state is the set of data stored in the satellite's memory. All satellites within the cluster share the global state, which is defined as follows:

[0109]

[0110] in, This indicates the state of the l-th satellite within the cluster. This represents the cluster state information collected by the star cluster. Indicates real-time business. This indicates non-real-time business.

[0111] (5) Action: As an intelligent agent, the satellite makes decisions based on its state at each moment, dynamically adjusts the beam scheduling of each time slot, and selects k cells from N cells for service, as defined below:

[0112]

[0113] in, N represents the beam scheduling status of the l-th satellite. S Indicates the total number of satellites; φ n =1 indicates that the nth cell was selected; N C This indicates the total number of cells; K represents the number of beams that the satellite can provide.

[0114] (6) Reward: Each agent receives an immediate reward after performing an action. To maximize the long-term optimization goal of the system, this invention uses three metrics to provide reward value for the agents:

[0115] First, the latency reward for real-time data packets is defined as follows:

[0116]

[0117] in Indicates the latency of data packets. This represents the total latency of all real-time data packets within the coverage area of ​​a single satellite.

[0118] Secondly, in order to maximize the throughput of non-real-time data packets, time slot t i The total amount of non-real-time data packets leaked is used as its throughput reward, defined as follows:

[0119]

[0120] Finally, in order to reduce interference between satellite beams, it is necessary to select a beam pattern with spatial isolation advantages from all beam scheduling schemes.

[0121] Define the adjacency matrix for each cell. Among the elements h ij ∈{0,1} represents cell c i and community c j Whether they are adjacent, h ij =1 indicates that the two communities are adjacent, and vice versa; the formula is as follows:

[0122]

[0123] Wherein d(c i ,c j ) represents cell c i and community c j The distance between them, where R represents the cell radius.

[0124] Define binary vector x i ∈{0,1} is used to represent the hopping beam pattern of a single satellite, x i =1 indicates cell c i Subject to beam service. When the service beams of a single satellite all meet the space isolation condition, there should be x. T Hx = 0. Therefore, the spatial isolation value is used as the interference avoidance reward, defined as follows:

[0125]

[0126] Because interference with satellite beams can be affected by the scheduling of other satellite beams, all satellites within the cluster, i.e., agents, share this reward.

[0127] By feeding the state and actions into the Q network for training, a beam-hopping scheme that meets the QoS requirements of different users and avoids inter-beam interference can be selected.

[0128] This invention comprises the following three parts:

[0129] (1) Inter-satellite load balancing strategy

[0130] Load balancing strategies for large-scale, multi-layered low-Earth orbit satellite constellations can be divided into inter-cluster load balancing and intra-cluster load balancing.

[0131] Inter-cluster load balancing refers to dividing the ground into several partitions of different sizes with similar traffic volumes based on service differences, and performing regional joint clustering of satellite constellations. By using edge satellite borrowing strategies, a balance between regional service density and the number of satellites can be achieved.

[0132] Intra-cluster load balancing targets the ground cells and satellites within a specific cluster, and determines the ground cells served by each satellite based on load balancing and latency requirements, thereby minimizing the load differences between satellites as much as possible.

[0133] (2) Multi-star beam skipping algorithm

[0134] This deep learning-based collaborative multi-agent beam hopping algorithm considers beam hopping resource scheduling within a satellite cluster. Each satellite is treated as an independent agent responsible for beam scheduling in its corresponding serving cell. While making independent decisions, each agent can share the global state and reward. In this stage, the multi-beam satellite acts as the decision-maker, using deep Q-learning to design beam hopping patterns based on the allocated cell's service requirements and the current onboard resources. This dynamically schedules resources by establishing and training a DQN network, satisfying as many needs as possible with limited beam resources, thus solving the beam hopping problem in multi-satellite collaboration. The DQN network will manage the state s... t As input and output action a t After the action is performed, the state is updated to s t+1 And you can get a reward r t Intelligent decision-making for beam scheduling is achieved through training a Q-network.

[0135] (3) Intra-cluster multi-satellite multi-beam load balancing

[0136] Since multi-layer LEO satellites typically have overlapping coverage areas, this invention requires further determination of the service cell for each satellite to maximize the utilization of multi-satellite resources within the cluster. First, each cell is determined based on the geographical relationship between the satellite and the cell. Visual satellite collection Vi And determine the cell within the star cluster. Belongs to the initial star cluster Ω m Whether a cell is located at the edge or center of the cluster, the edge cell within the cluster will have priority in selecting serving satellites. If the number of visible satellites in an edge cell is V... i If there is an increase compared to before clustering, then, considering indicators such as load balancing and latency requirements, priority will be given to selecting borrowed low-level low-Earth orbit (LEO) satellites or those within the original cluster; if the satellite cluster is sparse and satellites within the area have been borrowed, then the cell will select visible high-level LEO satellites for supplementary coverage. If the cell... If it belongs to the central cell and the star cluster belongs to the dense star cluster, then the cell... When selecting satellites, load balancing strategies are combined to select existing low-level low-Earth orbit satellites or high-level low-Earth orbit satellites as supplementary coverage; if the satellite cluster is a sparse cluster, existing low-level low-Earth orbit satellites are selected based on load balancing and the service classification (latency requirements) within the cell.

Claims

1. A cooperative beam-hopping method suitable for large-scale multi-layer low-Earth orbit satellite constellations, characterized in that, The specific steps are as follows: Step 1: Divide the Earth's surface evenly into areas S, and obtain the communication resource request size γ in each area, using a threshold γ... threshold As a zoning standard for business-intensive and business-sparse areas; Step 2: Obtain the satellite set within each partition based on the mapping relationship of the satellite sub-satellite points, which serves as the initial satellite cluster corresponding to that partition; Step 3: Calculate the ratio Δ of the number of satellites to the service density in each partition corresponding to each initial cluster, and sort the service-intensive partitions in ascending order according to Δ; for the service-intensive partitions, borrow edge satellites from adjacent areas in turn until the load of all partitions is balanced to form the final service cluster; The ratio Δ of the number of satellites to the service density is calculated using the following formula: Where k is the number of business-intensive or business-sparse partitions formed by merging the basic partitions, and N s The number of serving satellites corresponds to a region of area S; D is the service density of the region. The satellite secondment process is as follows: First, for the service-intensive partition with the smallest current satellite count to service density ratio Δ Select the service-sparse partition with the largest Δ among its adjacent partitions. And divided into business-sparse partitions Towards business-intensive partitions Provide satellites; Then, after each satellite is borrowed, update the number of satellites and the Δ table for the current service-intensive and service-sparse partitions; select the service-intensive partition with the smallest updated Δ. Borrowing the business sparse partition with the largest Δ among adjacent partitions The satellites in the data will be seconded, and the Δ table will be updated again. Until the regional load balancing of all partitions is satisfied The final service cluster can then be obtained; Indicates the threshold; Step 4: For each serving satellite cluster, select the respective serving satellites within the cluster for each cell based on the service classification of the cells within its coverage area; The intra-cluster cell selection process is as follows: First, regarding the current community C n The visible satellite set V corresponding to the cell is determined based on the geographical relationship between the satellite and the cell. n ; Then, determine cell c. n Whether a cell belongs to the edge or center of a zone is determined by the edge cell's priority in selecting the serving satellite. For edge cells: If the community belongs to a service-intensive zone, the number of visible satellites is V. i Compared to before the satellite borrowing, there is an increase, with priority given to borrowed low-level low-orbit satellites or original low-level low-orbit satellites within the cluster; by selecting the satellite with the least load, load balancing and time delay requirements of each satellite are achieved. If a cell belongs to a service-sparse zone and satellites within that zone are borrowed, then in cells without low-Earth orbit satellite coverage, visible high-Earth orbit satellites will be selected to supplement coverage, meaning that high-Earth orbit satellites will provide services to that cell. For the central residential area: If the cell belongs to a service-sparse zone, then the original visible low-level low-orbit satellite will be selected based on load balancing and service classification within the cell. If the community belongs to a service-intensive zone, load balancing strategies can be combined when selecting satellites to select existing low-level low-Earth orbit satellites or high-level low-Earth orbit satellites as enhanced supplementary coverage. This process continues until all cells have been visited and the serving satellites for all cells have been determined. Step 5: Treat each satellite in the cluster as an intelligent agent, responsible for beam scheduling of the corresponding serving cell. Use a deep DQN network to learn and coordinate the beam hopping resources of the satellites in the cluster, and finally select a beam hopping scheme that meets the QoS requirements of different users and avoids inter-beam interference. In the deep DQN network learning process (1) The state is defined as follows: in, This indicates the state of the l-th satellite within the cluster. This represents the cluster state information collected by the star cluster. Indicates real-time business. Indicates non-real-time business; (2) The action is defined as follows: in, N represents the beam scheduling status of the l-th satellite. S Indicates the total number of satellites; φ n =1 indicates that the nth cell was selected; N C This indicates the total number of cells; K represents the number of beams the satellite can provide. (3) Rewards: Three indicators are used to provide reward value for the agent; First, the latency reward for real-time data packets is defined as follows: in Indicates the latency of data packets. This represents the total latency of all real-time data packets within the coverage area of ​​a single satellite; Secondly, using time slot t i The total amount of non-real-time data packets leaked is used as its throughput reward, defined as follows: Indicates time slot t i satellite beam pattern, Indicates time slot t i Channel capacity of the beam-serving cell; Finally, the spatial isolation value, shared by all satellites within the cluster, is defined as the interference avoidance reward as follows: H represents the adjacency matrix of each cell. x i ∈{0,1} represents the binary vector of the hopping beam pattern of a single satellite, x i =1 indicates cell c i Subject to beam service; when the service beams of a single satellite all meet the space isolation condition, there should be x T Hx = 0.

2. The cooperative beam-hopping method for large-scale multi-layer low-Earth orbit satellite constellations as described in claim 1, characterized in that, In step one, the communication resource request γ≥γ threshold The areas designated as business-intensive areas are further divided into separate zones. The communication resource request γ < γ threshold Regions with sparse business operations are called business-sparse regions. These regions are selected from neighboring regions of the same type and merged to form new partitions. The specific number of neighboring regions of the same type to be merged is determined by a threshold γ. threshold The decision is made when the total communication resource requests of the sparse service area and all neighboring areas of the same type exceed the threshold γ. threshold This means the merger is complete; After several iterations, the ground communication area was divided into zones of varying sizes.

3. The cooperative beam-hopping method for large-scale multi-layer low-Earth orbit satellite constellations as described in claim 1, characterized in that, Step two specifically involves: For the m-th partition, if it is a business-intensive partition, its area is the same as that of the basic partition, i.e., S. m =S, Number of serving satellites N m =N s ; If it is a sparse business partition formed by merging k basic partitions, then the area S m =kS, the number of serving satellites satisfies N m =kN s ; Where, N m This indicates the number of service satellites corresponding to partition m.

Images