Multi-satellite cooperative hopping beam resource allocation method and system based on interference avoidance

Abstract

The invention relates to the technical field of satellite internet, and discloses a multi-satellite cooperative beam hopping resource allocation method and system based on interference avoidance, and the method comprises the steps: 1, constructing a multi-satellite beam hopping cooperative coverage interference analysis model; 2, modeling is carried out on a multi-satellite-hopping-beam multi-dimensional resource scheduling problem; 3, defining interference between spatial isolation measurement hopping beam patterns, constructing an interference constraint matrix, determining a shortest distance threshold value through a carrier-to-interference ratio threshold value, deducing an interference model in a same-rail scene and a different-rail scene, and bringing the carrier-to-interference ratio threshold value into optimization problem constraint; step 4, carrying out graph structure modeling based on GAT-PPO; 5, an Actor decision layer outputs a beam scheduling probability matrix and a power distribution vector in a double-branch mode, an action space is subjected to multiple constraints, and a reward function balances throughput and interference suppression; step 6, adopting a centralized training-distributed execution framework; according to the method, interference model analysis of multi-satellite cooperative coverage based on interference avoidance is constructed, and real-time analysis and scheduling of the interference condition are carried out.

Description

Technical Field

[0001] This invention relates to the field of satellite internet technology, specifically to a multi-satellite cooperative beam-hopping resource allocation method and system based on interference avoidance. Background Technology

[0002] Existing satellite communication interference analysis techniques have significant limitations in terms of dynamic adaptability: their modeling methods are mainly based on static orbital parameters, which cannot effectively characterize the time-varying interference characteristics under beam skipping scheduling scenarios; especially in multi-satellite collaborative coverage scenarios, existing technologies do not fully consider the spatiotemporal correlation of satellite trajectories and lack accurate descriptions of the dynamic evolution of interference caused by rapid beam skipping; this static modeling approach leads to bias in the identification of interference-sensitive areas, which seriously restricts the anti-interference performance of beam scheduling algorithms. Current AI scheduling technology faces fundamental challenges in multi-satellite collaborative scenarios: on the one hand, traditional algorithms ignore the inter-satellite interference coupling effect, resulting in serious resource allocation conflicts in overlapping coverage areas; on the other hand, as the number of satellites increases, the state space expands exponentially, causing the algorithm's convergence speed to drop sharply. This efficiency bottleneck makes it difficult to meet the real-time scheduling requirements of highly dynamic environments, and resource utilization efficiency is significantly limited. Summary of the Invention

[0003] The purpose of this invention is to solve the aforementioned problems by designing a multi-satellite cooperative beam-hopping resource allocation method and system based on interference avoidance. In low-Earth orbit (LEO) multi-satellite cooperative coverage systems, the uneven spatial distribution of ground users, rapid satellite movement, and rapid beam hopping lead to time-varying dynamic topology changes and uneven traffic demands. This has attracted more attention to advanced resource scheduling that meets heterogeneous and time-varying traffic demands. Furthermore, overlapping satellite coverage can cause strong co-channel interference due to sidelobe leakage, significantly impacting system performance. Efficient resource scheduling among different satellites is needed, and the impact of interference on communication quality in dynamically changing network topologies must be reduced. However, existing beam-hopping scheduling methods mostly focus on single satellites, rarely considering multi-satellite cooperative scheduling. Therefore, this invention addresses the problem of cooperative scheduling of overlapping coverage by co-orbiting and hetero-orbiting satellites in LEO multi-satellite beam-hopping systems. It conducts research on inter-satellite cooperative beam-hopping technology, constructs an interference model analysis based on interference avoidance for multi-satellite cooperative coverage, and performs real-time analysis and scheduling of interference.

[0004] The first aspect of this invention provides a multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance, the method comprising the following steps: Step 1: Construct a multi-satellite hopping beam cooperative coverage interference analysis model; Step 2: Model the multi-satellite hopping beam multi-dimensional resource scheduling problem, generate service requirements based on user Poisson point process distribution, and establish an optimization problem with the goal of maximizing throughput; Step 3: Define the spatial isolation measure for interference between hopping beam patterns, construct the interference constraint matrix, determine the shortest distance threshold through the carrier-to-interference ratio threshold, derive the interference model for both same-track and different-track scenarios, and incorporate the carrier-to-interference ratio threshold into the optimization problem constraints; Step 4: Based on GAT-PPO graph structure modeling, the satellite network is modeled as a topology graph. Nodes contain at least satellite location and link information. Edge weights are designed based on beam distance and interference intensity. The adjacency matrix is ​​dynamically updated as the satellite moves. Step 5: The GAT-PPO neural network is divided into a GAT network, a feature fusion layer, an Actor decision layer, and a Critic evaluation layer. The Actor decision layer has two branches that output a beam scheduling probability matrix and a power allocation vector. The action space is subject to multiple constraints, and the reward function balances throughput and interference suppression. Step 6: Adopt a centralized training-distributed execution framework. The Actor network optimizes based on experience pool data, constrains the update magnitude through a pruning function, and the Critic network minimizes the value estimation error, iterating until the cumulative reward is maximized.

[0005] Optionally, in the first implementation of the first aspect of the present invention, in the scenario of overlapping coverage by multiple satellites in different orbits in step 1, the ground area is modeled as a circular wave position, and the inter-satellite and intra-satellite interference is analyzed. The sidelobe interference intensity of the inter-satellite interference is calculated by equivalent omnidirectional radiation power, path loss, etc., and the intra-satellite interference is statistically analyzed by the co-frequency interference of different beams of the same satellite. Finally, the total interference intensity, carrier-to-interference ratio and signal-to-interference-to-noise ratio are obtained, and then the communication capacity is calculated based on the Shannon formula.

[0006] Optionally, in a second implementation of the first aspect of the present invention, the constraints in step 2 include the number of single-satellite beams, total power, communication capacity not exceeding demand, carrier-to-interference ratio threshold, and single-satellite scheduling limit.

[0007] Optionally, in a third implementation of the first aspect of the present invention, step 2 specifically includes: The user distributes the data across various wave positions according to a Poisson point process. The parameter of the Poisson point process distribution is density. Total number of users Follows a Poisson distribution, with parameters Each user's location is generated in a distribution manner, and each user is assigned to the nearest wave position. The number of users in each wave position is counted. For each user, their traffic requirements are generated based on the multi-satellite hop beam coordinated coverage interference analysis model; Each user periodically generates file transfer requests, with file arrival time intervals following an exponential distribution. The average session arrival rate is set as a parameter in the multi-satellite hopping beam cooperative coverage interference analysis model. Furthermore, the file size transmitted each time follows an exponential distribution. The service demand for the wave position can be obtained as follows: The system throughput is defined as the total amount of data successfully transmitted by all satellites to users at each frequency position within the scheduling period. The objective optimization problem for maximizing the total throughput is: in, For binary beam scheduling variables, representing satellites In the time slot Is the wave position scheduled? . It is a power allocation variable, representing the satellite In the time slot The transmission power; This indicates that each satellite can be scheduled at most once in a single time slot. One beam; This indicates that the total transmission power of the satellite must meet the maximum power limit; Indicates each wave position communication capacity Not exceeding its total demand ; This means that, to ensure communication quality, the carrier-to-interference ratio (CTR) of each scheduled bit must meet a minimum threshold. ; This means that to avoid resource conflicts, only one satellite is allowed to be scheduled within the same time slot for the same wavelength.

[0008] Optionally, in a fourth implementation of the first aspect of the present invention, step 3 specifically includes: use To measure spatial isolation and beam hopping pattern intervals (BHP) and BHP Spatial isolation is expressed as follows: in, Indicates wave position Center and wave position Distance from the center and They represent BHP respectively and BHP The number of the wave positions served are respectively . This reflects the shortest distance between two BHP service positions. The larger the value, the better the spatial interference isolation between BHPs; Define BHP disturbance constraints This represents the degree of mutual interference between two different satellites' BHP (Browser-Head-Off) parameters, and thus the satellite... and satellite BHP interference constraint matrix Represented as: By merging them, the BHP interference constraint matrix among all satellites can be obtained. Represented as: The spatial isolation metric between hop beam patterns (BHP) for each satellite is determined using an interference constraint matrix.

[0009] Optionally, in the fifth implementation of the first aspect of the present invention, the shortest distance threshold between BHP serving beam positions of the spatial isolation metric hopping beam pattern is used in different scenarios. The difference lies in the load-to-dryness ratio threshold. Determine the shortest distance threshold .

[0010] Optionally, in the sixth implementation of the first aspect of the present invention, step 4 specifically includes: In the satellite node, each satellite Position coordinates Determined by its ephemeris data, the dynamic changes in its coverage area need to be updated in real time through the ephemeris table. Due to the high-speed motion of low-Earth orbit satellites, the dynamic evolution of nodes over time needs to be modeled: Satellite movement causes downlink information coverage issues. The time shift is calculated from the transmit antenna gain, the real-time position coordinates of the satellite and the wave position, and the satellite's transmit antenna power. The signal power between the satellite and the wavelet, and the transmission requirements of the wavelet, are recalculated using the channel model. User traffic demand modeling is completed by generating a multi-satellite hopping beam collaborative coverage interference analysis model.

[0011] Optionally, in the seventh implementation of the first aspect of the present invention, the GAT-PPO algorithm in step 4 utilizes the beam distance between satellites and the relationship between adjacent beam numbers to design the edge. If the satellites and satellite The number of adjacent waveforms simultaneously scheduled between them is greater than 0, and the waveforms Sum of Waves Distance between Then the edges between the two satellite nodes are connected; Calculate the inter-satellite interference power received by the beam in a multi-satellite co-orbit scenario. The interference intensity is represented by the value of time-varying interference. In the case of multiple satellites with different orbits, the expected value of time-varying interference is calculated. As edge weight.

[0012] Optionally, in the eighth implementation of the first aspect of the invention, step 6 uses... To initialize the Actor network, each satellite in a time slot According to the current strategy Generate Actions Obtaining the current state from the environment And generate an adjacency matrix by constructing a topology graph. To characterize the network structure, the state Input Actor network to obtain and based on 'Dynamically select the optimal beam scheduling matrix and power allocation vector, based on the selected action' Interact with the environment and calculate rewards Next state The generated scheduling strategy is placed in the experience cache pool to optimize the Actor network.

[0013] A second aspect of the present invention provides a multi-satellite cooperative beam-hopping resource allocation system based on interference avoidance, the system comprising: The first building module is used to build a multi-satellite hopping beam cooperative coverage interference analysis model; The first modeling module is used to model the multi-star hopping beam multi-dimensional resource scheduling problem. It generates business requirements based on the user Poisson point process distribution and establishes an optimization problem with the goal of maximizing throughput. The second construction module is used to define the spatial isolation degree measure of interference between hopping beam patterns, construct the interference constraint matrix, determine the shortest distance threshold through the carrier-to-interference ratio threshold, derive the interference model for both same-track and different-track scenarios, and incorporate the carrier-to-interference ratio threshold into the optimization problem constraint. The second modeling module is used for graph structure modeling based on GAT-PPO. It models the satellite network as a topology graph. The nodes contain at least satellite position and link information. The weights of the edges are designed according to the wavelet distance and interference intensity. The adjacency matrix is ​​dynamically updated as the satellite moves. The output module, which is a neural network for GAT-PPO, consists of a GAT network, a feature fusion layer, an Actor decision layer, and a Critic evaluation layer. The Actor decision layer has two branches that output a beam scheduling probability matrix and a power allocation vector. The action space is subject to multiple constraints, and the reward function balances throughput and interference suppression. The iterative module employs a centralized training-distributed execution framework. The Actor network optimizes its strategy based on experience pool data and constrains the update magnitude through a pruning function. The Critic network minimizes the value estimation error and iterates until the cumulative reward is maximized.

[0014] The beneficial effects of the technical solution of this invention mainly include: 1. This invention addresses the core challenges of severe inter-satellite co-frequency interference and dynamically heterogeneous resource demands in low-Earth orbit multi-satellite cooperative beam-hopping scenarios. It conducts joint optimization of interference modeling and resource scheduling under multi-satellite overlapping coverage. Based on the actual characteristics of high-speed satellite motion and rapid beam hopping, and comprehensively considering inter-satellite interference coupling, beam position service demand distribution, and power constraints, a space-time-frequency multi-dimensional interference constraint matrix and dynamic isolation model are constructed. The evolution of interference links is accurately characterized by real-time updating of the adjacency matrix. Experiments show that in a 12-satellite cooperative scenario, this scheme reduces interference power to 3.1 × 10⁻⁶. -7 Below W, the average SINR reaches 14.6 dB, achieving a significant improvement of 42% in the avoidance rate of interference-sensitive areas; 2. This invention addresses the bottleneck problem of slow convergence and large optimization space in traditional algorithms for multi-satellite collaborative scheduling scenarios. It proposes a joint resource allocation design based on GAT and PPO. Based on the graph structure characteristics of the satellite network topology, it deeply integrates node locations, link states, and interference characteristics to design a dual-branch Actor network that synchronously generates beam scheduling probability matrices and power allocation vectors. The update magnitude of the strategy is constrained by a pruning function. Simulation results show that this scheme improves system throughput by more than 30% compared to traditional OSPF, PPO, and GDDR algorithms, with a single decision time of <50ms, achieving maximum resource utilization and a breakthrough improvement in scheduling efficiency under high dynamic environments. 3. In the innovative design of dynamic interference modeling and constraint integration, a time-varying interference model is constructed for multi-satellite collaborative beam-hopping scenarios. First, the inter-satellite and intra-satellite interference intensity is accurately quantified through spatial isolation measurement and interference constraint matrix. Second, the shortest distance threshold is dynamically derived based on the CIR threshold, and the interference constraint is embedded into the resource optimization problem, providing an innovative constraint mechanism for resource scheduling in the interference dimension under multi-satellite collaborative scenarios. 4. In the innovative design of joint resource allocation driven by GAT-PPO, a deep fusion architecture of graph neural network and reinforcement learning is proposed. First, GAT is used to dynamically extract interference features between satellite nodes. Second, a dual-branch Actor network is designed to output the beam scheduling probability matrix and power allocation vector. The policy update magnitude is constrained by the pruning function of PPO to achieve a balance between throughput improvement and interference suppression, and a complete resource allocation strategy from feature extraction to policy optimization is constructed. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 A flowchart of a multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance provided in an embodiment of the present invention; Figure 2 A schematic diagram of a low-orbit multi-satellite cooperative coverage system model provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a multi-satellite cooperative beam-hopping resource allocation system based on interference avoidance provided in an embodiment of the present invention. Detailed Implementation

[0017] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] The key issues to be addressed are as follows: (1) Dynamic interference modeling and interference suppression issues caused by multi-satellite overlapping coverage. In low Earth orbit (LEO) constellation systems, multiple satellites can simultaneously cover ground users in the same area. Especially in scenarios of overlapping coverage in the same or different orbits, sidelobe leakage of inter-satellite beams can cause severe co-channel interference and inter-satellite interference, seriously affecting system capacity and communication quality. Due to the highly dynamic nature of satellite orbital motion and beam hopping behavior, existing interference management methods are difficult to accurately characterize and respond to in real time. To this end, this invention considers a dynamic interference model for multi-satellite collaborative coverage in actual low Earth orbit (LEO) systems. Through spatial isolation measurement and interference constraint matrix, the intensity of inter-satellite and intra-satellite interference is accurately quantified. Considering inter-satellite interference and time-varying inter-satellite resources, an objective function that maximizes throughput is constructed.

[0019] (2) Resource scheduling and communication optimization in multi-satellite collaborative beam-hopping scenarios. In low-Earth orbit satellite constellation systems, the distribution of ground users is highly uneven. Coupled with the constellation's rapid mobility and beam-hopping behavior, the network faces highly dynamic and heterogeneous traffic demands. In such environments, traditional scheduling algorithms struggle to balance service quality and resource utilization efficiency. Especially in the context of multi-satellite collaborative scheduling, the differences in task requirements and priorities among different nodes further exacerbate the complexity of scheduling. To address these issues, this invention proposes a joint resource allocation algorithm based on a multi-satellite collaborative coverage interference model and a graph attention network and near-end policy optimization (GAT-PPO). By modeling the satellite network as graph-structured data, GAT dynamically extracts interference features between nodes, and the PPO algorithm is introduced to improve the stability of policy updates.

[0020] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 The flowchart of the multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance provided in this embodiment of the invention includes the following steps: Step 1: Construction of multi-satellite hopping beam cooperative coverage interference analysis model. Taking the overlapping coverage of multiple satellites with different orbits as an example, the ground area is modeled as a circular wave position. Inter-satellite and intra-satellite interference are analyzed. Inter-satellite interference is calculated by sidelobe interference intensity through equivalent omnidirectional radiated power and path loss. Intra-satellite interference is calculated by statistically analyzing the co-frequency interference of different beams of the same satellite. Finally, the total interference intensity, carrier-to-interference ratio (CIR) and signal-to-interference-to-noise ratio (SINR) are obtained. Then, the communication capacity is calculated based on Shannon's formula. Step 2: Modeling the multi-satellite hopping beam pattern multi-dimensional resource scheduling problem. Based on the user Poisson point process (PPP) distribution, service requirements are generated. An optimization problem is established with the goal of maximizing throughput. Constraints include the number of beams per satellite, total power, communication capacity not exceeding demand, CIR threshold, and single-satellite scheduling limit for each beam position. Step 3: Long-term multi-satellite collaborative coverage interference model analysis. Spatial isolation metric is defined to measure interference between hopping beam patterns (BHP). An interference constraint matrix is ​​constructed. The shortest distance threshold is determined through the CIR threshold. Interference models are derived for both same-track and different-track scenarios. The CIR threshold is incorporated into the optimization problem constraints. Step 4: Based on GAT-PPO graph structure modeling, the satellite network is modeled as a topology graph. Nodes contain states such as satellite position and link information. Edges are weighted according to the wavelet distance and interference intensity. The adjacency matrix is ​​dynamically updated as the satellite moves. Step 5: Design of GAT-PPO dynamic allocation strategy. The neural network includes GAT, feature fusion, Actor and Critic layers. The Actor has two branches that output beam scheduling probability matrix and power allocation vector. The action space is subject to multiple constraints. The reward function balances throughput and interference suppression. Step 6: Training process and optimization. A centralized training-distributed execution (CTDE) framework is adopted. The Actor network is optimized based on the experience pool data, and the update magnitude is constrained by the pruning function. The Critic network minimizes the value estimation error and iterates until the cumulative reward is maximized.

[0021] The specific implementation process for each step is given below: Step 1: (1-1) Construct a low-Earth orbit multi-satellite collaborative coverage system model like Figure 2 As shown, considering a low-Earth orbit multi-satellite cooperative beam-hopping system model, taking a scenario of overlapping coverage by multiple satellites in different orbits as an example, the ground area covered by the overlapping coverage by multiple satellites in different orbits is modeled as multiple circular wavefronts, and represented as... Each circular band can only be served by one LEO satellite beam at any given time. Each satellite provides coverage communication for the band via a spot beam. Assume each satellite has a maximum of [number missing] [satellites / satellites]. Each beam can be scheduled, and the beam is represented as... Due to the principle of circular wave position modeling, multiple wave positions may overlap at their edges. When adjacent satellites simultaneously schedule the same wave position, full-frequency reuse can cause significant interference.

[0022] (1-2) Constructing a multi-satellite hopping beam cooperative coverage interference analysis model make Representative satellite The first service One wave position, Represents being in a wave position Users receive from satellite Its wave position The interference signal strength of the scheduling beam, among which, The user received more than just satellite signals. Wave position Effective signal strength of the scheduling beam It will also receive signals from satellites. Schedule other wave positions Beam interference .

[0023] First, consider the interference between different satellites. Indicates satellite Its wave position The side lobes of the spot beam for satellite wave position The expression for co-channel interference generated by users is: in, It is path loss; This is the user's receiving antenna gain, which is a constant value when using an omnidirectional antenna; Indicates satellite to wave position The free space loss is for ,in, It is the operating frequency. Representative satellite to wave position The distance. Let be the equivalent isotropic radiated power of the satellite transmitting antenna, expressed as: This refers to the transmission power of the satellite antenna. This refers to the line loss of the antenna. It is a satellite Beam main lobe pointing position Side lobes point to wave position The angle between the main lobe and the lateral lobes; Indicates satellite The transmit antenna gain.

[0024] Assuming a satellite The geographic coordinates are Its spatial rectangular coordinates Similarly, the wave position can be calculated. coordinates Therefore, the interference angle can be calculated using the following formula: Simultaneously, wave positions can be further calculated. With satellite distance : Next, when considering the interference between different service beams of the same satellite, it can be understood as the satellite Equivalent satellite ,Right now At this time It is a satellite Wave position service beam sidelobe to wave position Co-channel interference generated by users, among which It is a satellite Beam main lobe pointing position Side lobes point to wave position The angle between the main lobe and the side lobes can be calculated similarly. . Indicates satellite Its wave position The sidelobe of the spot beam corresponds to the wave position The expression for co-channel interference generated by users is: Next, we will count the satellites. Intra-satellite beam position The intensity of co-channel interference needs to be determined based on the satellite's current position. The hopping beam pattern is used to determine the number of adjacent co-frequency interfering beams. The hopping beam scheduling period in the hopping beam system. The smallest granularity is the time slot, expressed as... In each time slot Each satellite has Each beam can be scheduled. Let Representative at the Satellites in each time slot wave position Is it subject to beam scheduling? If so, ,otherwise .make Representative at the Wave position within each time slot Received from satellite The expression for the interference intensity of other beams at the same frequency is: As can be seen, the interference value With satellite The hopping beam pattern is related. Similarly, let Representative at the Satellites within a time slot Service waveband The interference received from all other satellites' co-frequency beams is expressed as: This interference value is related to The value of is related. Therefore, the wave position The total interference intensity received is: Then define wave position The CIR is: Furthermore wave position Chinese users of for: Define wave position The number of users in Based on Shannon's formula, the wave position... communication capacity for: Step 2: Model the multi-satellite hopping beam multi-dimensional resource scheduling problem. Based on the system model and interference analysis, establish a throughput maximization objective function.

[0025] Users are distributed across various wavelengths according to the PPP distribution. The parameter of the PPP distribution is density. (Number of users per unit area), Total number of users Follows a Poisson distribution, with parameters Each user's location is generated according to the distribution described above, and each user is assigned to the nearest wave position. The number of users in each wave position is counted. For each user, their traffic requirements are generated according to the FTP-1 model. Each user will periodically generate file transfer requests, and the file arrival time interval follows an exponential distribution. The FTP-1 parameter, average session arrival rate, is set. Furthermore, the file size transmitted each time follows an exponential distribution. The service demand for the wave position can be obtained as follows: We define system throughput as the total amount of data successfully transmitted by all satellites to users at each frequency position within the scheduling period. Therefore, the optimization problem for maximizing total throughput is: in, For binary beam scheduling variables, representing satellites In the time slot Is the wave position scheduled? . It is a power allocation variable, representing the satellite In the time slot The transmission power. Among them, This indicates that each satellite can be scheduled at most once in a single time slot. One beam; This indicates that the total transmission power of the satellite must meet the maximum power limit; Indicates each wave position communication capacity Not exceeding its total demand ; This means that, to ensure communication quality, the CIR of each scheduled bit must meet a minimum threshold. ; This means that to avoid resource conflicts, only one satellite is allowed to be scheduled within the same time slot for the same wavelength.

[0026] Step 3: Severe inter-satellite interference may occur between BHPs of different satellites. It is necessary to ensure that the BHPs of different satellites within the same time slot meet spatial isolation requirements as much as possible. Therefore, we define... To measure BHP and BHP Spatial isolation is expressed as follows: in, Indicates wave position Center and wave position Distance from the center and They represent BHP respectively and BHP The number of the wave positions served are respectively . This reflects the shortest distance between two BHP service positions. The larger the value, the better the spatial interference isolation between BHPs. Therefore, BHP interference constraints are defined. This represents the degree of mutual interference between two different satellites' BHP (Browser-Head-Off) parameters, and thus the satellite... and satellite BHP interference constraint matrix Represented as: By merging them, the BHP interference constraint matrix among all satellites can be obtained. Represented as: The BHP of each satellite is determined using an interference constraint matrix, but the shortest distance threshold for the BHP serving spectral position varies depending on the scenario. Unlike others, we use CIR thresholds. Determine the shortest distance threshold Considering the interference models for binary stars in the same orbit and binary stars in different orbits, for the scenario of multiple stars in the same orbit, the CIR threshold can be calculated using the following lemma. .

[0027] Lemma: In the scenario of two LEO satellites in the same orbit, the satellites operate at the Ka band, the satellite transmitting antenna is a circular aperture reflector antenna, the satellite coverage radius is R, the beam radius is r, and the user elevation angle is not considered. The two satellites are fixed directly above the center wave position. The carrier-to-interference ratio reaches its maximum value when the distance between the wave position and the center line in the overlapping area of ​​the two satellites is d.

[0028] The interference threshold constraint for CIR is: Considering the interference between multiple satellites, how to design a resource allocation scheme for multi-satellite coordinated scheduling to maximize bandwidth throughput? Therefore, the above problem can be modeled as the following optimization problem: Step 4: (4-1) Model the complex multi-satellite cooperative coverage interference scenario as a topology diagram, starting with node design.

[0029] In the satellite node, each satellite Position coordinates Determined by its ephemeris data, the dynamic changes in its coverage area need to be updated in real time via the ephemeris table. Due to the high-speed motion of low-Earth orbit satellites, the dynamic evolution of nodes over time requires accurate modeling. Satellite movement causes downlink information coverage issues The time shift is calculated from the transmit antenna gain, the real-time position coordinates of the satellite and the wave position, and the satellite's transmit antenna power. The signal power between the satellite and the wavelet is recalculated using the channel model. The transmission requirements of the wavelet are also considered. User traffic demand modeling is completed using the FTP-1 model. Traffic demand can be modeled as follows: in For the set of users within the wave position, For users Packet arrival rate, This is the packet size. Therefore, the satellite node... Carrying status information The specific information included is as follows: (4-2) Next, we will design the edges.

[0030] The GAT-PPO algorithm can design edge relationships by utilizing the distance between simultaneously scheduled satellite positions and the number of adjacent positions. However, overlapping satellite positions are prone to co-channel interference when simultaneously scheduled. Therefore, it is determined that if satellites... and satellite The number of adjacent waveforms simultaneously scheduled between them is greater than 0, and the waveforms Sum of waves Distance between If the two satellite nodes are connected by an edge, then the interference distance threshold is [not specified]. The interference model from the previous section can be used for calculation. The inter-satellite interference power received by the beam in a multi-satellite scenario with the same orbit can be calculated using the lemma. The interference intensity is represented by the value of time-varying interference. In the case of multiple satellites with different orbits, the expected value of time-varying interference is calculated. As edge weights. Specifically, the storage structure of the state graph can use an adjacency matrix. To express, Represents a node and nodes The weights of the connecting edges are as follows: Due to the high-speed movement of satellites, the adjacency matrix It is necessary to recalculate for each time slot, and simultaneously iterate through all satellites to calculate the sum of interference power of adjacent positions, dynamically adjusting the side weights. This strengthens the attention allocation to high-interference links, ensuring that interference-sensitive areas pointed out by the lemma are actively avoided during policy generation.

[0031] Step 5: (5-1) Design a dynamic allocation strategy for multi-satellite cooperative beam hopping resources based on GAT-PPO. First, design a neural network.

[0032] The GAT-PPO neural network consists of four parts: the GAT network, the feature fusion layer, the Actor decision layer, and the Critic evaluation layer. For the multi-satellite collaborative coverage beam hopping scheduling problem, the input to the GAT is the state information of all satellites. .

[0033] Among them, satellite eigenvectors It includes downlink information, service requirement information, and allocated power information; and , The state and adjacency matrix When input into the GAT network, the feature vectors of each node are first linearly transformed through a fully connected layer. The number of neurons in the fully connected layer is related to the dimension of the node's feature vector. The same. The multi-head graph attention layer has 4 heads, each with a hidden dimension of 32, and the output, after concatenation, has 128 dimensions, hence the new state. The dimension is 256. The attention coefficient for each head is calculated as follows: in For attention vectors, To share the weight matrix, the feature vectors of adjacent nodes are finally processed according to the attention coefficients. Perform a weighted summation to complete the fused feature vector of the current node. Then, the new state is obtained. .

[0034] Then, the state As input to the Actor network. The number of neurons in the input layer of the deep neural network (DNN) of the Actor network is the same as the output data of the GAT network. The dimensions are the same, followed by fully connected layers with 128, 64, and 32 neurons respectively, and ReLU activation function. Since this paper has two scheduling variables, one is the beam scheduling variable... and power allocation variables Therefore, the Actor network is designed with a two-branch output. The first branch is the beam scheduling decision branch, with the number of neurons in the output layer set to 32 and 1. The scheduling probability is generated through the Sigmoid activation function. This leads to the output of the wave position scheduling probability matrix. Next is the power allocation decision branch, where the number of neurons in the output layer is set to 32 and 1, and mapped to the power normalization value through the sigmoid activation function. Then, inverse normalization is performed: Finally, output the power allocation vector. The Critic network is also composed of a DNN network, and its input is environmental state information, i.e., 123-dimensional fused features. The number of neurons in the input layer is... The fused features of all satellite nodes are stitched together to form a global state, and the output is the state value through a fully connected layer design. The output layer has 1 neuron, representing the current state value. Simultaneously, the CIR threshold and the dynamic threshold for out-of-track interference derived from the lemma are introduced. As a state value correction term. When node The real-time CIR value is lower than At this time, the Critic network imposes a nonlinear penalty on the value estimation, forcing the policy update to prioritize satisfying the disturbance constraint.

[0035] (5-2) Next, the motion space is designed.

[0036] The action space is defined as a joint decision, where For binary beam scheduling variables, representing satellites In the time slot Is the wave position scheduled? . It is a power allocation variable, representing the satellite In the time slot The transmit power. In the beam scheduling decision branch, lemmas and inter-orbit interference constraints need to be considered to reduce the dimensionality of the action space. Co-orbit satellites and Generate a binary mask matrix based on the threshold of the lemma. Simultaneous beam scheduling of satellites in the same orbit within the critical distance is prohibited; satellites in different orbits... and Define the maximum number of beams that can be scheduled for satellites in different orbits: ,in For system interference capacity, This is background noise. (Through constraints) Avoid combinations with high interference.

[0037] After outputting the transition pattern, the GAT-PPO algorithm still needs to allocate the beam power vector. The Actor network outputs the beam scheduling probability matrix for each satellite, which is specific to each satellite. Each action probability value is calculated. Then, based on a greedy strategy, the matrix with the largest probability value is selected as the hopping beam pattern. Similarly, the power allocation vector is output and inversely normalized to the power allocation value for each beam. Therefore, the satellite... Action space Wave position scheduling probability matrix and power allocation vector The composition can be represented as: Low-Earth orbit multi-satellite cooperative beam-hopping scenario It consists of 10 satellites, so the action space of its entire system can be represented as: Meanwhile, the motion space must meet the following constraints: Single-satellite beam count constraint: per satellite In the time slot Maximum allocation within One beam: Waveband scheduling constraint: Each waveband can only be scheduled by one satellite. Total power constraint: The beam power of each satellite cannot exceed the total power. (5-3) Finally, the reward function is designed.

[0038] The reward function needs to balance maximizing throughput and suppressing interference, and specifically consists of the following parts: defining a reward for meeting demand, encouraging satellites to meet beam requirements as much as possible, with the reward being proportional to the amount of data transmitted: in For satellite wavelength service demand, The amount of service data allocated to satellites based on wavebands. An interference penalty term is defined based on lemmas and extra-orbit interference constraints to penalize high-interference scenarios: The first penalty term originates from the CIR threshold constraint of the lemma, and the second term is based on the dynamic model of inter-track interference to suppress interference in overlapping regions. This represents the co-frequency interference value experienced by the wavefront from adjacent satellites in the same orbit. The interference threshold set for this system can be obtained from the interference model analysis in the previous section. Define the comprehensive reward function: in and These are the weighting coefficients, which are then optimized through simulation.

[0039] Step 6: Use the CTDE framework for training and optimization. First, use... To initialize the Actor network, each satellite in a time slot According to the current strategy Generate Actions Then obtain the current state from the environment. And generate an adjacency matrix by constructing a topology graph. To characterize the network structure; then the state Input Actor network to obtain and based on Dynamically select the optimal beam scheduling matrix and power allocation vector. Based on the selected action... Interact with the environment and calculate rewards Next state The generated scheduling strategy is placed in the experience cache pool, which allows for the optimization of the Actor network. The optimization objective of the Actor network is: in, and These represent the old and new policy parameters, respectively. This represents the number of all scheduling policies in the experience cache pool. This indicates the time interval for selecting the scheduling strategy. Defined as: During the strategy optimization process and These correspond to the policy parameterization distributions before and after the iterative update; the advantage function. The Critic network is used to calculate the relative gain of the current action compared to the historical average performance. To constrain the policy update magnitude, a clipping function is introduced, limiting the probability ratio of the new and old policies to a range. Within, the hyperparameters The pruning boundary is controlled and is typically set to 0.1-0.3 to balance exploration and stability. The training objective of the Critic network is to minimize the value estimation error, and its loss function is defined as: In the formula, This is the estimated value of the state value function. As a discount factor, and , Indicates the first Instant reward for each step. Network parameters are updated via gradient descent. This allows the Critic network to asymptotically approximate the expected distribution of the true cumulative reward. The table below provides the pseudocode for the multi-star cooperative scheduling algorithm based on the GAT-PPO algorithm.

[0040] Table 1. Pseudocode for the implementation of the multi-star cooperative scheduling algorithm based on GAT-PPO Please see Figure 3 A schematic diagram of the structure of a multi-satellite cooperative beam-hopping resource allocation system based on interference avoidance provided in this embodiment of the invention. The system includes: The first building module is used to build a multi-satellite hopping beam cooperative coverage interference analysis model; The first modeling module is used to model the multi-star hopping beam multi-dimensional resource scheduling problem. It generates business requirements based on the user Poisson point process distribution and establishes an optimization problem with the goal of maximizing throughput. The second construction module is used to define the spatial isolation degree measure of interference between hopping beam patterns, construct the interference constraint matrix, determine the shortest distance threshold through the carrier-to-interference ratio threshold, derive the interference model for both same-track and different-track scenarios, and incorporate the carrier-to-interference ratio threshold into the optimization problem constraint. The second modeling module is used for graph structure modeling based on GAT-PPO. It models the satellite network as a topology graph. The nodes contain at least satellite position and link information. The weights of the edges are designed according to the wavelet distance and interference intensity. The adjacency matrix is ​​dynamically updated as the satellite moves. The output module, which is a neural network for GAT-PPO, consists of a GAT network, a feature fusion layer, an Actor decision layer, and a Critic evaluation layer. The Actor decision layer has two branches that output a beam scheduling probability matrix and a power allocation vector. The action space is subject to multiple constraints, and the reward function balances throughput and interference suppression. The iterative module employs a centralized training-distributed execution framework. The Actor network optimizes its strategy based on experience pool data and constrains the update magnitude through a pruning function. The Critic network minimizes the value estimation error and iterates until the cumulative reward is maximized.

[0041] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance, characterized in that, The method includes the following steps: Step 1: Construct a multi-satellite hopping beam cooperative coverage interference analysis model; Step 2: Model the multi-satellite hopping beam multi-dimensional resource scheduling problem, generate service requirements based on user Poisson point process distribution, and establish an optimization problem with the goal of maximizing throughput; Step 3: Define the spatial isolation measure for interference between hopping beam patterns, construct the interference constraint matrix, determine the shortest distance threshold through the carrier-to-interference ratio threshold, derive the interference model for both same-track and different-track scenarios, and incorporate the carrier-to-interference ratio threshold into the optimization problem constraints; Step 4: Based on GAT-PPO graph structure modeling, the satellite network is modeled as a topology graph. Nodes contain at least satellite location and link information. Edge weights are designed based on beam distance and interference intensity. The adjacency matrix is ​​dynamically updated as the satellite moves. Step 5: The GAT-PPO neural network is divided into a GAT network, a feature fusion layer, an Actor decision layer, and a Critic evaluation layer. The Actor decision layer has two branches that output a beam scheduling probability matrix and a power allocation vector. The action space is subject to multiple constraints, and the reward function balances throughput and interference suppression. Step 6: Adopt a centralized training-distributed execution framework. The Actor network optimizes based on experience pool data, constrains the update magnitude through a pruning function, and the Critic network minimizes the value estimation error, iterating until the cumulative reward is maximized.

2. The multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance as described in claim 1, characterized in that, In step 1, for the scenario of multiple satellites overlapping and covering different orbits, the ground area is modeled as a circular wave position. Inter-satellite and intra-satellite interference are analyzed. The sidelobe interference intensity of inter-satellite interference is calculated by equivalent omnidirectional radiated power and path loss. The intra-satellite interference is statistically analyzed by co-frequency interference of different beams of the same satellite. Finally, the total interference intensity, carrier-to-interference ratio and signal-to-interference-to-noise ratio are obtained, and then the communication capacity is calculated based on Shannon's formula.

3. The multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance as described in claim 1, characterized in that, The constraints in step 2 include the number of beams per satellite, total power, communication capacity not exceeding demand, carrier-to-interference ratio threshold, and single-satellite scheduling limits.

4. The multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance as described in claim 1, characterized in that, Step 2 includes the following specific steps: The user distributes the data across various wave positions according to a Poisson point process. The parameter of the Poisson point process distribution is density. Total number of users Follows a Poisson distribution, with parameters Each user's location is generated in a distribution manner, and each user is assigned to the nearest wave position. The number of users in each wave position is counted. For each user, their traffic requirements are generated based on the multi-satellite hop beam coordinated coverage interference analysis model; Each user periodically generates file transfer requests, with file arrival time intervals following an exponential distribution. The average session arrival rate is set as a parameter in the multi-satellite hopping beam cooperative coverage interference analysis model. Furthermore, the file size transmitted each time follows an exponential distribution. The service demand for the wave position can be obtained as follows: The system throughput is defined as the total amount of data successfully transmitted by all satellites to users at each frequency position within the scheduling period. The objective optimization problem for maximizing the total throughput is: in, For binary beam scheduling variables, representing satellites In the time slot Is the wave position scheduled? . It is a power allocation variable, representing the satellite In the time slot The transmission power; This indicates that each satellite can be scheduled at most once in a single time slot. One beam; This indicates that the total transmission power of the satellite must meet the maximum power limit; Indicates each wave position communication capacity Not exceeding its total demand ; This means that, to ensure communication quality, the carrier-to-interference ratio (CTR) of each scheduled bit must meet a minimum threshold. ; This means that to avoid resource conflicts, only one satellite is allowed to be scheduled within the same time slot for the same wavelength.

5. The multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance as described in claim 1, characterized in that, Step 3 includes the following specific steps: use To measure spatial isolation and beam hopping pattern intervals (BHP) and BHP Spatial isolation is expressed as follows: in, Indicates wave position Center and wave position Distance from the center and They represent BHP respectively and BHP The number of the wave positions served are respectively . This reflects the shortest distance between two BHP service positions. The larger the value, the better the spatial interference isolation between BHPs; Define BHP disturbance constraints This represents the degree of mutual interference between two different satellites' BHP (Browser-Head-Off) parameters, and thus the satellite... and satellite BHP interference constraint matrix Represented as: By merging them, the BHP interference constraint matrix among all satellites can be obtained. Represented as: The spatial isolation metric between hop beam patterns (BHP) for each satellite is determined using an interference constraint matrix.

6. The multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance as described in claim 5, characterized in that, Spatial isolation measures the shortest distance threshold between BHP serving beam positions between hopping beam patterns in different scenarios. The difference lies in the load-to-dryness ratio threshold. Determine the shortest distance threshold .

7. The multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance as described in claim 1, characterized in that, Step 4 includes the following specific steps: In the satellite node, each satellite Position coordinates Determined by its ephemeris data, the dynamic changes in its coverage area need to be updated in real time through the ephemeris table. Due to the high-speed motion of low-Earth orbit satellites, the dynamic evolution of nodes over time needs to be modeled: Satellite movement causes downlink information coverage issues. The time shift is calculated from the transmit antenna gain, the real-time position coordinates of the satellite and the wave position, and the satellite's transmit antenna power. The signal power between the satellite and the wavelet, and the transmission requirements of the wavelet, are recalculated using the channel model. User traffic demand modeling is completed by generating a multi-satellite hopping beam collaborative coverage interference analysis model.

8. The multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance as described in claim 1, characterized in that, In step 4, the GAT-PPO algorithm uses the beam distance between satellites and the number of adjacent beams to design the relationship between the edges. If the satellites and satellite The number of adjacent waveforms simultaneously scheduled between them is greater than 0, and the waveforms Sum of waves Distance between Then the edges between the two satellite nodes are connected; Calculate the inter-satellite interference power received by the beam in a multi-satellite co-orbit scenario. The interference intensity is represented by the value of time-varying interference. In the case of multiple satellites with different orbits, the expected value of time-varying interference is calculated. As edge weight.

9. The multi-satellite cooperative beam-hopping resource allocation method based on interference avoidance as described in claim 1, characterized in that, Step 6 uses To initialize the Actor network, each satellite in a time slot According to the current strategy Generate Actions Obtaining the current state from the environment And generate an adjacency matrix by constructing a topology graph. To characterize the network structure, the state Input Actor network to obtain and based on 'Dynamically select the optimal beam scheduling matrix and power allocation vector, based on the selected action' Interact with the environment and calculate rewards Next state The generated scheduling strategy is placed in the experience cache pool to optimize the Actor network.

10. A multi-satellite cooperative beam-hopping resource allocation system based on interference avoidance, characterized in that, The system includes: The first building module is used to build a multi-satellite hopping beam cooperative coverage interference analysis model; The first modeling module is used to model the multi-star hopping beam multi-dimensional resource scheduling problem. It generates business requirements based on the user Poisson point process distribution and establishes an optimization problem with the goal of maximizing throughput. The second construction module is used to define the spatial isolation measure of interference between hopping beam patterns, construct the interference constraint matrix, determine the shortest distance threshold through the carrier-to-interference ratio threshold, derive the interference model for both same-track and different-track scenarios, and incorporate the carrier-to-interference ratio threshold into the optimization problem constraint. The second modeling module is used for graph structure modeling based on GAT-PPO. It models the satellite network as a topology graph. The nodes contain at least satellite position and link information. The weights of the edges are designed according to the wavelet distance and interference intensity. The adjacency matrix is ​​dynamically updated as the satellite moves. The output module, which is a neural network for GAT-PPO, consists of a GAT network, a feature fusion layer, an Actor decision layer, and a Critic evaluation layer. The Actor decision layer has two branches that output a beam scheduling probability matrix and a power allocation vector. The action space is subject to multiple constraints, and the reward function balances throughput and interference suppression. The iterative module employs a centralized training-distributed execution framework. The Actor network optimizes its strategy based on experience pool data and constrains the update magnitude through a pruning function. The Critic network minimizes the value estimation error and iterates until the cumulative reward is maximized.

Images