Hybrid routing method and system under the condition of limited deployment of gateway stations in low-orbit mega constellation network

Abstract

This invention discloses a hybrid routing method and system for low-Earth orbit (LEO) mega-constellation networks under conditions of limited gateway station deployment. The method includes: Access satellite and gateway station selection: Based on the geographical location of ground users, the minimum hop count from the access satellite to each gateway station is calculated using a hop count evaluation model, and the combination of access satellites and gateway stations with the minimum hop count is selected; Multi-hop awareness routing decision: A multi-hop load-aware area is established along the transmission path from the access satellite to the gateway station. Paths are dynamically planned according to a link cost metric model, where the link cost is calculated by weighting queue occupancy, congestion probability, and bandwidth utilization, and congestion nodes are bypassed; Gateway station load balancing control: Within the gateway station coverage area, traffic is allocated using a backpressure routing strategy based on the queue backlog difference at satellite ports to achieve load balancing. This method achieves synergistic optimization in latency, congestion control, and load balancing, significantly enhancing network stability and service quality.

Description

Technical Field

[0001] This invention relates to the field of satellite communication technology, specifically to a hybrid routing method and system for low-Earth orbit mega-constellation networks under conditions of limited gateway deployment. Background Technology

[0002] Low Earth Orbit (LEO) mega-constellation systems are typically represented by Starlink. These systems consist of three main parts: a space segment, a ground segment, and a user segment. The space segment comprises over a thousand LEO satellites and inter-satellite links, forming the backbone network for space transmission. The ground segment consists of gateway stations, an integrated operations and control management system, and a ground core network. The user segment includes various user terminals, a comprehensive information service platform, and business support systems. Ground user terminals establish satellite-to-ground links with the satellites. After user data is transmitted to the satellites, it is forwarded via inter-satellite links and then transmitted down to the gateway stations via feeder links, from where it connects to the ground core network.

[0003] Because gateway station location selection is constrained by geographical, economic, and political factors, gateway stations are likely to be deployed only in a limited region or country rather than globally. With the increasing demand from mobile users and limited satellite buffers, uneven global population distribution will lead to uneven satellite load. Furthermore, this deployment method requires more hops or longer latency compared to global deployment. This forces routing algorithms to consider both load balancing for long-distance transmission and how to reduce the number of hops to lower latency. In particular, it is necessary to consider the more severe congestion problems caused by traffic concentration near gateway stations. Therefore, this deployment method will place more stringent requirements on routing algorithm design. Designing a stable and efficient routing mechanism is crucial for improving the performance of large-scale LEO constellation networks. Summary of the Invention

[0004] To address some or all of the technical problems existing in the prior art, this invention provides a hybrid routing method and system under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network.

[0005] The technical solution of the present invention is as follows:

[0006] Firstly, a hybrid routing method is provided under the condition of limited deployment of gateway stations in a low-Earth orbit mega-constellation network. This method includes:

[0007] Access satellite and gateway station selection: Based on the geographical location of the ground user, the minimum number of hops from the access satellite to each gateway station is calculated using a hop count evaluation model, and the combination of access satellite and gateway station with the minimum number of hops is selected;

[0008] Multi-hop aware routing decision: A multi-hop load awareness area is established in the transmission path from the access satellite to the gateway station. The path is dynamically planned according to the link cost measurement model. The link cost is calculated by weighting queue occupancy rate, congestion probability and bandwidth utilization, and the congested nodes are bypassed.

[0009] Gateway station load balancing control: Within the coverage area of ​​the gateway station, traffic is distributed using a backpressure routing strategy based on the queue backlog difference of satellite ports to achieve load balancing.

[0010] In an optional embodiment, the hop-count-based evaluation model includes:

[0011] Based on the latitude and longitude of the ground user and the ascent or descent type of the access satellite, the minimum hop count from the access satellite to each gateway station is calculated. The combination of the access satellite and gateway station with the minimum hop count is selected, where the hop count is determined by a weighted sum of the hop counts in the same orbital plane and the hop counts in different orbital planes.

[0012] In an optional embodiment, the link cost metric is calculated based on a combination of the following parameters:

[0013] The queue occupancy rate of a satellite node indicates the current buffer status of the satellite output port queue;

[0014] Queue congestion probability is used to predict the congestion status of satellite nodes by combining the base probability and trend factor through a sliding time window.

[0015] Link bandwidth utilization is dynamically assessed using a sliding time window and exponential moving average method to evaluate the real-time load of the link.

[0016] The link cost metric is obtained by weighted summation and applying a nonlinear penalty term, which causes the cost of high-load links to increase rapidly.

[0017] In an optional embodiment, the calculation of the queue congestion probability includes the following steps:

[0018] Construct a fixed-size sliding time window to record the queue occupancy rate sequence at several recent time points;

[0019] Calculate the base probability, which is determined based on the ratio of the current queue occupancy rate to a preset congestion threshold;

[0020] A trend factor is calculated by comparing the difference between the current queue occupancy rate and the average occupancy rate within the sliding time window, and then normalizing the difference.

[0021] The queue congestion probability is calculated by combining the basic probability and trend factor to predict the congestion status of satellite nodes.

[0022] In an optional embodiment, the calculation of the link bandwidth utilization includes the following steps:

[0023] Construct a fixed-size sliding time window, periodically sample the data transmitted through the link, and record the cumulative number of bytes transmitted at each sampling point;

[0024] Calculate the instantaneous bandwidth utilization rate, which is the ratio of the actual number of bytes transmitted within the window to the theoretical maximum transmission capacity;

[0025] The instantaneous bandwidth utilization is smoothed using the exponential moving average method to obtain the current bandwidth utilization. The smoothing factor is between 0 and 1 and is used to balance instantaneous responsiveness and stability.

[0026] In an optional embodiment, the construction of the multi-hop load-aware region includes:

[0027] Centered on the current satellite, periodically collect the congestion status and link cost of satellites within a range of n≥2 hops;

[0028] When the queue occupancy rate of a satellite node exceeds a threshold, a limited flooding update of the neighbor node status is triggered.

[0029] Within the sensing area, the Dijkstra algorithm is used to calculate the shortest path, prioritizing the edge nodes with the fewest hops and the lowest link cost.

[0030] In an optional embodiment, when multiple edge nodes have the same number of hops, a remaining path selection domain that is close to a square is selected to improve the path survival probability, where the path survival probability is defined as the probability of avoiding congested links.

[0031] In an optional embodiment, the backpressure routing strategy includes:

[0032] Construct an n≥2 hop sensing area centered on the gateway satellite, covering the ascending and descending orbit satellites around the gateway station;

[0033] Calculate the backlog difference between adjacent satellites based on the queue backlog metric of each satellite port;

[0034] Within the path area with the minimum number of hops, the next hop node with the largest backlog difference is selected for data packet transmission until it reaches the gateway station.

[0035] In an optional embodiment, the queue backlog metric is based on the independent queues of each output port of the satellite, defined as the backlog in a certain port direction, and the transmission rate allocation is optimized by the link cost metric.

[0036] Secondly, a hybrid routing system is provided under the condition of limited deployment of gateway stations in a low-Earth orbit mega-constellation network. This system includes:

[0037] The access selection module is configured to calculate the minimum hop count from the access satellite to each gateway station based on the geographical location of the ground user and the hop count evaluation model, and select the combination of access satellite and gateway station with the minimum hop count.

[0038] The multi-hop sensing routing module is configured to establish a multi-hop load sensing area in the transmission path from the access satellite to the gateway station, dynamically plan the path according to the link cost metric model, wherein the link cost is calculated by weighting queue occupancy rate, congestion probability and bandwidth utilization, and bypass congested nodes.

[0039] The load balancing module is configured to distribute traffic within the coverage area of ​​the gateway station using an improved backpressure routing strategy based on the queue backlog difference of the satellite ports, thereby achieving load balancing.

[0040] The main advantages of the technical solution of this invention are as follows:

[0041] This invention presents a hybrid routing method and system for low-Earth orbit (LEO) mega-constellation networks under constrained gateway station deployment conditions. In the access satellite and gateway station selection phase, based on minimum hop count evaluation, it prioritizes the combination of access satellites and gateway stations with optimal latency, effectively reducing end-to-end transmission hops and propagation latency. In the multi-hop-aware routing decision phase, it proactively avoids congested nodes and reduces packet loss rate through dynamic link cost metrics (integrating queue occupancy, congestion probability, and bandwidth utilization) and local path planning. In the gateway station load balancing control phase, it employs an improved backpressure routing strategy, dynamically allocating traffic based on the satellite port queue backlog difference to alleviate structurally high loads around gateway stations, thereby improving network throughput and data delivery rate. The overall solution achieves synergistic optimization in latency, congestion control, and load balancing, making it particularly suitable for scenarios with uneven traffic distribution caused by centralized gateway station deployment, significantly enhancing network stability and service quality. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart illustrating a hybrid routing method under constrained gateway station deployment conditions in a low-Earth orbit mega-constellation network according to an embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of the structure of a low-Earth orbit mega-constellation network according to an embodiment of the present invention;

[0045] Figure 3This is a schematic diagram of the satellite topology in a low-Earth orbit mega-constellation network according to an embodiment of the present invention;

[0046] Figure 4 This is a schematic diagram of the cache corresponding to each satellite port in a low-Earth orbit mega-constellation network according to an embodiment of the present invention;

[0047] Figure 5 This is a schematic diagram of a multi-hop path from an access satellite to a gateway satellite in a low-Earth orbit mega-constellation network according to an embodiment of the present invention;

[0048] Figure 6 This is a schematic diagram of a multi-hop load sensing region (n=3) in a low-Earth orbit mega-constellation network according to an embodiment of the present invention;

[0049] Figure 7 Satellite S in a low-Earth orbit mega-constellation network according to an embodiment of the present invention a A diagram illustrating the selection of the next-hop satellite within the minimum hop count path range;

[0050] Figure 8 This is an overall flowchart of a hybrid routing method under limited deployment conditions for gateway stations in a low-Earth orbit mega-constellation network according to an embodiment of the present invention;

[0051] Figure 9 This is an architecture diagram of a hybrid routing system under limited deployment conditions of gateway stations in a low-Earth orbit mega-constellation network according to an embodiment of the present invention;

[0052] Figure 10 This is a schematic diagram illustrating the variation of data delivery rate with data flow rate in a hybrid routing method under limited gateway station deployment conditions in a low-Earth orbit mega-constellation network according to an embodiment of the present invention.

[0053] Figure 11 This is a schematic diagram illustrating the variation of throughput with data flow rate in a hybrid routing method under limited gateway deployment conditions in a low-Earth orbit mega-constellation network according to an embodiment of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0055] The technical solutions provided by the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0056] Low Earth orbit (LEO) mega-constellation networks (such as Starlink) consist of a space segment (satellites and inter-satellite links), a ground segment (gate stations, etc.), and a user segment. Due to geographical, economic, or political factors, gate stations may be concentrated in specific areas (such as within China), leading to network congestion, uneven load distribution, and long-distance transmission latency issues.

[0057] To address the aforementioned problems, in a first aspect, embodiments of the present invention provide a hybrid routing method under conditions of limited gateway station deployment in a low-Earth orbit mega-constellation network, as shown in the appendix. Figure 1 As shown, it includes:

[0058] Step 1, Access Satellite and Gateway Station Selection: Based on the geographical location of the ground user, calculate the minimum number of hops from the access satellite to each gateway station using the hop count evaluation model, and select the combination of access satellite and gateway station with the minimum number of hops.

[0059] The purpose of this step is to calculate the minimum number of hops from the access satellite to each gateway station based on the geographical location of the ground user and using a hop count evaluation model, and to select the combination with the minimum number of hops to reduce propagation latency. Hop count refers to the number of relays required for data to travel from the access satellite to the gateway satellite (the satellite connecting the gateway station), including hops within the same orbital plane and hops across orbital planes.

[0060] For example, the specific steps include:

[0061] User location identification: The system obtains the latitude and longitude of the ground user, for example, the user is located in Shanghai (latitude 30°N, longitude 120°E).

[0062] Visible satellite identification: Based on the user's location and satellite orbital parameters (such as the Walker-Delta constellation's orbital altitude of 550km and inclination of 53°), calculate which satellites are currently within the user's field of view. Typically, satellites with an elevation angle greater than 10° are considered visible. Assuming two satellites are identified: S1 (ascending orbit) and S2 (descending orbit).

[0063] Hop Count Calculation: For each visible satellite, the system calculates the minimum hop count to the gateway satellites of each gateway station (e.g., Beijing, Shanghai, Guangzhou). The hop count is based on the constellation's "+Grid" topology (each satellite has 4 inter-satellite links: 2 in the same orbital plane and 2 across orbital planes), and is quickly looked up using a pre-stored topology table. For example: S1 to gateway satellite S10 at the Shanghai gateway station requires 2 hops in the same orbital plane + 1 hop across orbital plane, for a total of 3 hops. S2 to gateway satellite S10 at the Shanghai gateway station requires 1 hop in the same orbital plane + 1 hop across orbital plane, for a total of 2 hops.

[0064] Optimal combination selection: The system compares the hop count of all combinations and selects the combination with the smallest hop count. In this example, the hop count from S2 to the Shanghai gateway station is the smallest at 2, therefore S2 is selected as the access satellite and the Shanghai gateway station as the target. The selection process is completed by the embedded processor on the satellite, with a calculation time of less than 100 milliseconds, ensuring real-time performance.

[0065] Step 2, Multi-hop Aware Routing Decision: Establish a multi-hop load awareness area in the transmission path from the access satellite to the gateway station, dynamically plan the path according to the link cost metric model, the link cost is calculated by weighting queue occupancy rate, congestion probability and bandwidth utilization, and bypass congested nodes.

[0066] This step establishes a multi-hop load awareness zone along the transmission path from the access satellite to the gateway station. Based on a link cost metric model, it dynamically plans the path to bypass congested nodes, achieving low-latency transmission and congestion avoidance. The link cost metric comprehensively considers queue occupancy, congestion probability, and bandwidth utilization, determining the load level of each link through weighted calculations.

[0067] For example, the specific steps include:

[0068] Multi-hop load sensing area establishment: Centered on the current satellite (e.g., access satellite S2), construct an n-hop (n≥2) load sensing area. For example, when n=3, based on the "+Grid" topology, it covers approximately 24 satellites within a 3-hop range.

[0069] The system broadcasts via inter-satellite links, periodically (e.g., every second) collecting load information from satellites within the region, including queue occupancy (reflecting the current buffer status), congestion probability (predicting the likelihood of future congestion), and bandwidth utilization (the real-time load of the link). This information is stored in the local database of the current satellite, occupying approximately 1KB of memory.

[0070] Link cost metric calculation: The cost of each inter-satellite link is calculated by weighting three parameters to ensure that all factors are taken into account.

[0071] Path planning: Within the sensing area, the system constructs a directed graph where nodes are satellites, edges are inter-satellite links, and edge weights represent link costs. Dijkstra's algorithm is used to calculate the optimal path from the current satellite to the edge node (the satellite with the fewest hops from the gateway station), prioritizing low-cost (low-load) links. For example, from S2 to edge node S3, the path is S2→S6→S3, avoiding satellite S5 with high queue occupancy.

[0072] Dynamic update: Data packets are forwarded along the computing path. After reaching the edge node, the system reconstructs the sensing area with that node as the center and repeats the above process until the data enters the sensing area of ​​the gateway satellite.

[0073] Step 3, Gateway Station Load Balancing Control: Within the coverage area of ​​the gateway station, traffic is distributed using a backpressure routing strategy based on the queue backlog difference of satellite ports to achieve load balancing.

[0074] This step, within the coverage area of ​​the gateway station (i.e., near the gateway satellite), uses a backpressure routing strategy to distribute traffic based on the difference in queue backlog at the satellite ports, achieving load balancing and improving throughput and data delivery rate. Backpressure routing naturally achieves load balancing by directing data to neighboring nodes with less backlog. For example, the specific steps include:

[0075] Sensing Area Establishment: When data enters the n-hop (n≥2) sensing area of ​​a gateway satellite (e.g., S10), the system constructs a fixed sensing area centered on S10, covering surrounding ascending and descending orbit satellites. For example, when n=3, it covers approximately 24 satellites. Area information is collected via inter-satellite link broadcast, updated every 100 milliseconds.

[0076] Queue backlog measurement: Each satellite maintains independent queues for multiple output ports (corresponding to inter-satellite links), and the system measures the backlog at each port in real time (i.e., the amount of data waiting to be processed in the queue). For example, the backlog at port 1 of satellite S8 is Q. S 8 1 (t) = 1000 bytes, and the backlog at port 2 of its neighbor S9 is Q. S 9 2 (t) = 500 bytes.

[0077] Backlog Calculation: For the S8 to S9 link, calculate the backlog:

[0078]

[0079] Similarly, calculate the backlog difference between S8 and other neighbors (such as S13, S14), for example, 300 bytes and -200 bytes (negative values ​​indicate that S8's backlog is less than that of its neighbors).

[0080] Next-hop selection: The system selects the neighbor with the largest backlog difference as the next hop (i.e., the neighbor with the smallest backlog). In this example, the backlog difference between S8 and S9 is the largest (500 bytes), so the data is forwarded from S8 to S9. The selection process is completed on the processor of S8 and takes about 1 millisecond.

[0081] Continuous forwarding: Data packets are forwarded hop-by-hop until they reach the gateway satellite S10, and then transmitted to the gateway station. The transmission rate is allocated using a maximum weight matching algorithm to ensure efficient bandwidth utilization.

[0082] In summary, the hybrid routing method provided in this embodiment of the invention for low-Earth orbit mega-constellation networks under conditions of limited gateway station deployment prioritizes the optimal combination of access satellites and gateway stations in the minimum hop count evaluation phase, effectively reducing end-to-end transmission hops and propagation latency. In the multi-hop-aware routing decision phase, dynamic link cost metrics (integrating queue occupancy, congestion probability, and bandwidth utilization) and local path planning actively avoid congested nodes, reducing packet loss rate. In the gateway station load balancing control phase, an improved backpressure routing strategy dynamically allocates traffic based on the satellite port queue backlog difference, alleviating structurally high loads around gateway stations and improving network throughput and data delivery rate. The overall solution achieves synergistic optimization in latency, congestion control, and load balancing, and is particularly suitable for scenarios with uneven traffic distribution caused by centralized gateway station deployment, significantly enhancing network stability and service quality.

[0083] In some optional implementations of the embodiments of the present invention, the above-mentioned hop count-based evaluation model includes:

[0084] Based on the latitude and longitude of the ground user and the ascent or descent type of the access satellite, the minimum hop count from the access satellite to each gateway station is calculated. The combination of the access satellite and gateway station with the minimum hop count is selected, where the hop count is determined by a weighted sum of the hop counts in the same orbital plane and the hop counts in different orbital planes.

[0085] First, enter the user's location: the system obtains the latitude and longitude of the ground user, for example, a user in Shanghai (latitude 30°N, longitude 120°E).

[0086] Next, identify the visible satellites: Based on the satellite orbital parameters (Walker-Delta constellation, 550km altitude, 24 orbits × 22 satellites), calculate the visible satellites at the current moment, categorizing them into ascending (northward) and descending (southward) orbits. Assume that S1 (ascending orbit, elevation angle 15°) and S2 (descending orbit, elevation angle 12°) are identified.

[0087] Then, the hop count is calculated: The system is based on the constellation's "+Grid" topology and pre-stores a hop count table for each satellite to the gateway satellite. For example, the gateway satellite for the Shanghai Gateway Station is S10 (descending orbit). For S1 (ascending orbit) to S10: 2 hops within the same orbital plane (moving along the orbital plane to a satellite close to S10) and 1 hop to a different orbital plane (switching to the orbital plane of S10) are required, for a total of 3 hops. For S2 (descending orbit) to S10: 1 hop within the same orbital plane and 1 hop to a different orbital plane are required, for a total of 2 hops. The weighted summation formula is: J total =w intra ·J intra +w inter ·J inter .

[0088] Among them, w intra=0.6, indicating the same orbital plane weight, reflecting lower latency, w inter =0.4, indicating the weight of the different orbital planes, taking into account switching costs. J intra and J inter These represent the hop counts for the same orbital plane and the hop counts for different orbital planes, respectively. For example, the hop count for S2 is calculated as: J total =0.6·1+0.4·1=1.0.

[0089] Finally, the optimal combination is selected: the number of hops from S1 and S2 to each gateway station (Beijing, Shanghai, Guangzhou) is compared, and the number of hops from S2 to the Shanghai gateway station is the smallest, so S2 and the Shanghai gateway station are selected.

[0090] In this embodiment, the initial path selection is optimized by considering the type of ascending and descending orbits and the weighted hop count. The propagation delay is reduced by precise path selection, and access satellites with the same orbit type as the gateway satellites (such as descending orbit to descending orbit) are selected first to reduce orbital plane switching and reduce switching jitter.

[0091] In some optional implementations of this invention, the link cost metric is calculated based on the following parameters:

[0092] The queue occupancy rate of a satellite node indicates the current buffer status of the satellite output port queue;

[0093] Queue congestion probability is used to predict the congestion status of satellite nodes by combining the base probability and trend factor through a sliding time window.

[0094] Link bandwidth utilization is dynamically assessed using a sliding time window and exponential moving average method to evaluate the real-time load of the link.

[0095] The link cost metric uses a weighted summation and applies a non-linear penalty term, which causes the cost of high-load links to increase rapidly.

[0096] The queue occupancy rate is obtained as follows: Each satellite's output port (e.g., 4 inter-satellite link ports) maintains an independent queue, and the queue occupancy rate is defined as the ratio of the current cache size to the maximum cache capacity. For example, if satellite S6's port 1 cache is 800 bytes and its maximum capacity is 1000 bytes, the occupancy rate Q is... occupy =800 / 1000=0.8.

[0097] The probability of queue congestion is obtained as follows: Congestion likelihood is predicted using historical data. A sliding time window (e.g., the most recent 5 seconds) is used to record the queue occupancy sequence, combined with a base probability (based on the current occupancy rate) and a trend factor (the trend of occupancy rate changes). For example, if the occupancy rate of S6 continues to rise, the predicted congestion probability P is... congestion =0.7.

[0098] The link bandwidth utilization rate is obtained as follows: Data transmission is sampled within a sliding time window (e.g., 1 second), the ratio of actual transmission volume to maximum capacity is calculated, and smoothed using an exponential moving average (EMA). For example, if the link transmission rate from S6 to S7 is 8 Mbps and the maximum capacity is 10 Mbps, the utilization rate U... bandwidth =8 / 10=0.8.

[0099] The formula for calculating link cost using weighted summation and penalty is as follows:

[0100] C link =w1·Q occupy +w2·P congestion +w3·U bandwidth

[0101] Where w1, w2, and w3 can be 0.4, 0.3, and 0.3, respectively. If any parameter exceeds a threshold (e.g., 0.8), a non-linear penalty term is applied. For example, C link =C link ·(1+0.5·(Q occupy -0.8). For S6 to S7, assume Q occupy =0.8, P congestion =0.7, U bandwidth =0.8, then C link = 0.4·0.8 + 0.3·0.7 + 0.3·0.8 = 0.77. If Q occupy =0.9, the cost after penalty increases to about 1.0.

[0102] In this embodiment, the load is accurately assessed by combining three parameters, which can accurately reflect the link status and optimize path selection. The congestion probability prediction function can avoid potential congestion in advance and improve data delivery rate. Furthermore, a non-linear penalty term ensures that high-load links are effectively bypassed, enhancing network stability under high-load scenarios.

[0103] In some optional implementations of the embodiments of the present invention, the calculation of the above-mentioned queue congestion probability includes the following steps:

[0104] Construct a fixed-size sliding time window to record the queue occupancy rate sequence at a number of recent time points. For example, the window size can be set to 5 seconds, with sampling every 0.1 seconds, recording 50 queue occupancy rate data points. For instance, the occupancy rate sequence for port 1 of satellite S6 could be 0.6, 0.65, 0.7, ..., 0.8.

[0105] Calculate the base probability, which is determined based on the ratio of the current queue occupancy to a preset congestion threshold. For example, calculate the base probability based on the ratio of the current occupancy to a threshold (e.g., 0.8):

[0106]

[0107] Among them, P base Based on the probability, Q occupy Q represents the current queue occupancy rate. threshold This is a preset congestion threshold. If Q... occupy =0.8, Q threshold =0.8, then P base =0.8 / 0.8=1.0.

[0108] The trend factor is calculated by comparing the current queue occupancy rate with the average occupancy rate within the sliding time window, and then normalizing the difference. For example, the difference between the current occupancy rate and the average occupancy rate within the window is calculated as follows:

[0109]

[0110] in, Let's assume the average occupancy rate within the window. Q occupy =0.8, then ΔQ = 0.8 - 0.7 = 0.1. After normalization, the trend factor is obtained:

[0111]

[0112] Assuming the maximum difference is 0.2, then T trend =0.1 / 0.2 = 0.5.

[0113] By combining baseline probability and trend factors, a weighted calculation of queue congestion probability is performed to predict the congestion status of satellite nodes. For example, combining baseline probability and trend factors:

[0114] P congestion =w base ·P base +w trend ·T trend

[0115] Among them, w base It can be 0.6, w trend If it can be 0.4, then P congestion =0.6·1.0+0.4·0.5=0.8.

[0116] As can be seen, the above embodiments ensure that predictions are based on the latest data through time windows, adapting to changes in network load and improving the accuracy of path selection. Trend factors capture rising occupancy trends to predict congestion risk, and early warnings reduce packet loss rates. Furthermore, fixed windows and normalization processes reduce computational complexity, making them suitable for satellite embedded systems.

[0117] In some optional implementations of this invention, the calculation of link bandwidth utilization includes the following steps:

[0118] Construct a fixed-size sliding time window to periodically sample link transmission data and record the cumulative number of bytes transmitted at each sampling point. For example, set the window size to 1 second, sample once every 0.1 seconds, and record the cumulative number of bytes transmitted at 10 time points. For instance, the link transmission data from S6 to S7 might be 1MB, 1.2MB, ..., 0.8MB.

[0119] Calculate the instantaneous bandwidth utilization rate, which is the ratio of the actual number of bytes transmitted within the window to the theoretical maximum transmission capacity. The formula is as follows:

[0120]

[0121] Assuming the total transfer volume within the window is 10MB and the maximum capacity is 12.5MB (corresponding to 10Mbps, 1s), then U instant =10 / 12.5=0.8.

[0122] The instantaneous bandwidth utilization is smoothed using an exponential moving average method to obtain the current bandwidth utilization. The smoothing factor is between 0 and 1, used to balance instantaneous responsiveness and stability. For example, the formula for smoothing the instantaneous utilization is as follows:

[0123] U bandwidth (t)=α·Uinstant(t)+(1-α)·U bandwidth (t-1)

[0124] The smoothing factor α can be 0.3, balancing responsiveness and stability. If U in the previous time step... bandwidth (t-1)=0.75, current U instant If (t) = 0.8, then U bandwidth (t) = 0.765.

[0125] In this embodiment, the instantaneous bandwidth utilization is smoothed using the Exponential Moving Average (EMA) method to reduce the impact of instantaneous fluctuations, improve the stability of bandwidth utilization assessment, optimize path selection, and the EMA algorithm is simple and suitable for satellite resource-constrained environments. A sliding time window and a suitable smoothing factor ensure rapid adaptation to load changes and reduce path latency.

[0126] In some optional implementations of the embodiments of the present invention, the construction of the multi-hop load-aware region includes:

[0127] Centered on the current satellite, the system periodically collects the congestion status and link costs of satellites within a range of n ≥ 2 hops. For example, centered on the current satellite (e.g., S2), an n = 3-hop sensing area is constructed, covering 24 satellites. Congestion status and link costs are collected every second via inter-satellite link broadcast.

[0128] When the queue occupancy rate of a satellite node exceeds a threshold, a limited flooding is triggered to update the state of neighboring nodes. For example, if the queue occupancy rate threshold is set to 0.8, and the port occupancy rate of S2 reaches 0.9, a limited flooding is triggered to broadcast the updated state to neighbors within 3 hops. The broadcast range is limited to 3 hops to avoid excessive network overhead.

[0129] Within the sensing area, Dijkstra's algorithm is used to calculate the shortest path, prioritizing the edge node with the fewest hops and lowest link cost. For example, within the sensing area, a directed graph is constructed, and Dijkstra's algorithm is run based on link cost to calculate the shortest path to the edge node. The edge node with the fewest hops to the gateway satellite (e.g., S3, 2 hops) is prioritized; if the hop counts are the same, the node with the lowest link cost is selected.

[0130] In this embodiment, periodic and triggered updates ensure the real-time nature of load information, reducing congested path selection. Limited flooding restricts the broadcast range, reducing communication overhead. Dijkstra's algorithm, combined with hop count and cost optimization, can reduce path latency.

[0131] In some optional implementations of this invention, when multiple edge nodes have the same hop count, a remaining path selection domain that is close to square is selected to improve the path survival probability, where the path survival probability is defined as the probability of avoiding congested links. For example, within a 3-hop sensing area, assume that the hop count from edge nodes S3 and S4 to gateway satellite S10 is both 2. The remaining path selection domains (the range of subsequently available paths) from S3 and S4 to S10 are calculated. The domain of S3 is close to square (covering satellites in four directions), while the domain of S4 is relatively long and narrow (mainly along a single orbital plane). The square domain is determined through geometric analysis; for example, the domain of S3 contains 8 selectable satellites, and the domain of S4 contains 5. In this case, S3 is preferred because it has more alternative paths, which can improve the path survival probability.

[0132] In some optional implementations of the embodiments of the present invention, the above-mentioned backpressure routing strategy includes:

[0133] A sensing region of n≥2 hops is constructed centered on the gateway satellite, covering both ascending and descending orbit satellites surrounding the gateway station. For example, a sensing region of n=3 hops is constructed centered on gateway satellite S10, covering 24 ascending and descending orbit satellites. Payload information is broadcast every 100 milliseconds.

[0134] Based on the queue backlog metric for each satellite port, the backlog difference between adjacent satellites is calculated. For example, measuring the satellite port backlog, such as port 1 of S8 having a backlog of 1000 bytes and the corresponding port 2 of S9 having a backlog of 500 bytes, the backlog difference is:

[0135]

[0136] Within the minimum hop count path area, the next-hop node with the largest backlog difference is selected for data packet transmission until it reaches the gateway station. For example, within the minimum hop count path area (hop count to S10 ≤ 3), the neighbor with the largest backlog difference (such as S9) is selected, and data is forwarded hop by hop to S10.

[0137] In this embodiment, data flow is guided to low-load satellites by the backlog difference, optimizing traffic allocation and reducing load imbalance.

[0138] In some optional implementations of this invention, the queue backlog metric is defined as the backlog amount in a certain port direction, based on the independent queue of each satellite's output port, and the transmission rate allocation is optimized using the link cost metric. For example, each satellite maintains four output port queues, and the backlog amount is the amount of data to be transmitted at each port. For instance, the backlog amount at port 1 of S8 is 1000 bytes. Optimizing the transmission rate allocation using the link cost metric means allocating higher rates to lower-cost links. For example, the link cost from S8 to S9 is 0.77, allocating 80% bandwidth; the cost to S13 is 1.0, allocating 20% ​​bandwidth. The rate is determined using a maximum weight matching algorithm.

[0139] The following describes the hybrid routing method under the condition of limited gateway station deployment in the low-Earth orbit mega-constellation network of the present invention with reference to another specific embodiment.

[0140] The embodiments of the present invention are applied to the following: Figure 2 The diagram shows a model of a low-Earth orbit mega-constellation system. Ground users include fixed ground terminals, airborne terminals, shipborne terminals, and handheld mobile terminals. Satellites that establish connections with ground users are called Access Satellites (AS), and satellites that establish connections with gateway stations are called Gateway Satellites (GS).

[0141] After ground users send data packets to the access satellite via the user link, the data packets are transmitted between satellites via inter-satellite links according to the routing policy. The inter-satellite links adopt a "+Grid" configuration, meaning each satellite has four inter-satellite links: two within the same orbital plane and two outside the same orbital plane. The length of the within-the-orbit inter-satellite links remains constant, while the length of the outside-the-orbit inter-satellite links varies. After the data packets are transmitted to the gateway satellite, they are sent to the gateway station via the feeder link, and the gateway station connects to the Internet. To describe the satellite congestion state, it is assumed that each output port of the satellite has a queue to buffer data packets to be forwarded.

[0142] The constellation adopts a Walker-Delta configuration, assuming it contains M×N satellites, where M is the number of orbital planes and N is the number of satellites per orbital plane. Represents the m-th satellite in the n-th orbital plane, and the set of all satellites is as follows: If the orbital inclination of each orbital plane is θ, then the constellation can cover a latitude range of [-θ, θ]. Satellite S is located within the orbital plane. m,n When its latitude value rate of change with time t When, it is defined as a satellite ascending to orbit; when At that time, it is defined as a de-orbiting satellite. Assume there are W gateway stations. Represents the w-th gateway, and the set of all gateways is... The collection of all gateway satellites is It is important to note that the satellites in the gateway satellite set will change as they switch with the gateway station. A gateway station can typically connect to multiple satellites as gateway satellites. This invention assumes that the gateway station simultaneously establishes feed links with at least one ascending satellite and one descending satellite at a single orbital altitude. Furthermore, it assumes that the number of gateway stations is small and they are only deployed within China. Unlike polar orbit constellations, the Walker-Delta configuration has no gaps between adjacent orbits; therefore, the Walker-Delta constellation network can be simplified into a 2D-Torus network, such as... Figure 3 As shown.

[0143] To facilitate understanding of the technical solution of this invention, the following explanations will first cover queue congestion probability prediction, link bandwidth utilization, and link cost measurement.

[0144] Queue congestion probability prediction:

[0145] Because the satellite has four inter-satellite links, each port has a queue, such as Figure 4 As shown, the size of the cache in each queue is defined as follows: Then the overall queue occupancy rate of the satellite node is:

[0146]

[0147] Among them, Q max This represents the maximum length that the queue can buffer. The queue occupancy rate of a satellite node's port is defined as:

[0148]

[0149] Where x represents {1,2,3,4}, which represent the four port directions: up, down, left, and right.

[0150] To accurately and quickly predict satellite node congestion, and since node congestion is directly related to satellite queue occupancy, this invention designs a satellite congestion prediction model based on queue occupancy. This model predicts the probability of network congestion through sliding window observation, basic probability calculation, and trend analysis. The model always maintains a queue of size N. q A sliding time window records the most recent N... q The queue occupancy rate at each time point forms a sequence W. q (t):

[0151] W q (t)={O x (tN q +1), O x (tN q +2),...,O x (t)} (3)

[0152] Congestion probability P of congestion prediction model congestion From the basic probability P base and trend factor F trend It consists of two parts. The base probability reflects the impact of the current queue state on congestion, and the formula for calculating the base probability is:

[0153]

[0154] Where, θ c This is the congestion threshold. The trend factor reflects the trend of queue growth and can be obtained by calculating the average occupancy rate within a sliding time window. The calculation formula is:

[0155]

[0156] Therefore, the trend value T(t) at the current time point is:

[0157]

[0158] This represents the difference between the current queue occupancy rate and the average occupancy rate of the sliding window. The normalized trend factor is obtained as follows:

[0159]

[0160] Where, θ t It is a trend threshold, used to normalize trend values.

[0161] Finally, taking into account both the base probability and the trend factor, the congestion probability is obtained, calculated using the following formula:

[0162] P congestion (t)=min(1,P base (t)+F trend (t)) (8)

[0163] As can be seen, when the queue occupancy rate is below the congestion threshold, the base probability is 0, and congestion prediction is mainly determined by the trend factor. When the queue occupancy rate is close to the queue capacity, the base probability is close to 1, and the system issues a high congestion warning. When the queue is in a rapid growth phase, if the trend value is positive, the trend factor provides additional early warning, helping to detect potential congestion risks in advance. During the rapid growth phase of the queue, the trend factor can provide early warning, helping to detect potential congestion risks in advance. The congestion probability prediction method has advantages such as adaptability, early warning capability, adjustability, and high computational efficiency. It can dynamically capture changes in network state, balance response and stability, and limit computational complexity.

[0164] Link bandwidth utilization:

[0165] To accurately measure the real-time load status of a link, this invention designs a bandwidth utilization calculation method based on a sliding time window. This method achieves dynamic evaluation of bandwidth utilization by continuously monitoring link data transmission and combining it with an exponential moving average.

[0166] First, a fixed-size sliding time window is constructed to collect data transmitted over the link. Each sampling point within the window records the cumulative number of bytes transmitted. The system samples periodically at a preset update interval, adding new data points to the end of the window while removing expired sampling points that are outside the window's time range, thus maintaining a dynamically updated data sequence. The model always maintains a window of size N. l A sliding time window records the most recent N... l The queue occupancy rate at each time point forms a sequence W. l (t):

[0167] W l (t)={B tx (tN l +1), B tx (tN l +2),...,B tx (t)} (9)

[0168] Among them, B tx (t) represents the number of bytes cumulatively transmitted at time t on this port.

[0169] Secondly, in each update cycle, the actual number of bytes transmitted within the window is calculated. By subtracting the cumulative number of bytes transmitted from the first and last sampling points of the window, the actual data transmission volume within that time period can be obtained:

[0170] B real =B tx (t)-B tx (tN l +1) (10)

[0171] Simultaneously, based on the link's nominal data transmission rate and window duration, the theoretical maximum transmission capacity B is calculated. max Instantaneous bandwidth utilization is the ratio of the actual number of bytes transmitted to the theoretical maximum capacity.

[0172]

[0173] To reduce the impact of network fluctuations on measurement results, this invention employs the Exponential Moving Average (EMA) method to smooth the instantaneous bandwidth utilization. Specifically, the current bandwidth utilization is calculated using the following formula:

[0174] U(t) = α·U instant (t)+(1-α)·U instant (t-1) (12)

[0175] Where U(t) represents the bandwidth utilization at time t, U instant (t) represents the instantaneous utilization rate calculated so far, and α is a smoothing factor (0 < α < 1). By adjusting the value of α, a balance can be achieved between instantaneous responsiveness and stability: a larger value of α makes the system more sensitive to changes in network load, while a smaller value of α provides more stable measurement results.

[0176] Link cost metric:

[0177] In order to ensure that routing decisions can reflect the real-time status of the current network and also have a certain degree of foresight, a comprehensive evaluation index is designed based on the previously derived evaluation indicators such as queue occupancy rate, queue congestion probability, and link bandwidth utilization to measure the link cost between satellite nodes.

[0178] Define the link cost metric function L x (t):

[0179]

[0180] Where ω1, ω2, and ω3 are weighting coefficients, and ω1 + ω2 + ω3 = 1. The satellite node occupancy rate O(t) reflects the current satellite congestion status, and the congestion probability P... congestion U(t) reflects the probability of congestion in the queue within a future period, and U(t) reflects the current link utilization. These metrics are weighted and summed to obtain a comprehensive measure of the link's state, which is then multiplied by... The term can measure values ​​that are more sensitive to congestion states, when O x (t) exceeds the congestion threshold θ c hour, The number of terms will increase rapidly, and k is a penalty term that will further accelerate this process. The term increases the speed, therefore L x The lower the value, the higher the priority of the link.

[0181] The main advantage of designing a link cost metric function is that it adopts a multi-dimensional comprehensive evaluation mechanism. It not only considers the queue occupancy and bandwidth utilization of the current network state, but also introduces congestion probability to predict future trends, realizing non-linear penalties for high-load links. At the same time, it uses adjustable weight coefficients to balance the importance of various indicators, ensuring that routing decisions can reflect the real-time status of the current network and have a certain degree of foresight. This design helps to achieve more balanced and efficient load distribution in scenarios where the network topology changes dynamically.

[0182] This invention divides the routing process from ground users to gateway stations into three stages. Stage I: Ground user access satellite and ground gateway station selection stage. This stage is used to determine the ground user's access satellite and the final gateway station where the route will land. By estimating the number of hops required from the access satellite to each gateway station, the optimal access satellite is found to ensure that the entire route maintains low latency and jitter. Stage II: Routing from the access satellite to the vicinity of the gateway satellite. A semi-distributed multi-hop awareness strategy is proposed. Using the link cost metric model proposed in the previous section, the locally optimal path is found in each multi-hop awareness area to achieve low propagation latency and congestion avoidance. Stage III: Gateway station-centered load balancing strategy. This stage is mainly used to perform load balancing on the satellites around the gateway satellite of the gateway station. An improved backpressure routing strategy is adopted to increase the total throughput of the gateway station.

[0183] Phase I, Selection of Access Satellites and Gateway Stations:

[0184] To reduce propagation delay during routing, data packets typically land at nearby gateway stations to access the terrestrial network. However, the straight-line distance from a satellite to a gateway station does not accurately reflect the propagation delay. This is because the unique configuration of satellite constellations allows data packets to propagate only via inter-satellite links. The establishment rules for these links are constellation parameters (such as phase factors), and even nearby gateway stations may take detours, resulting in higher propagation delays. Therefore, using the "hop count" of data packets during inter-satellite transmission is a more representative way to assess propagation delay.

[0185] Studies have shown that the hop count between different satellite types (ascending or descending orbits) is different and can be evaluated. Therefore, the actual routing process begins with the selection of the access satellite. Assume each ground gateway station can connect to at least one ascending orbit satellite and one descending orbit satellite. In the scenario of a large LEO constellation, there are also many satellites within the line of sight of ground users that can serve as access satellites. Assuming that ground users can also choose the ascending or descending orbit satellite closest to them as their access satellite, then based on the latitude, longitude, and ascending / descending orbit type of the access satellite, the minimum hop count for each gateway station can be estimated, and the access satellite and gateway station with the minimum hop count can be selected for routing. The packet propagation delay from the access satellite (AS) to the gateway satellite (GS), as the routing cost, is a key indicator for handover and ensures minimal end-to-end latency.

[0186] The satellite's nadir point (SSP) trajectory can be expressed in latitude. And longitude λ represents, for example Figure 5 As shown. Note the latitude. The orbital inclination must be maintained within the range of θ. u∈[-π,π] represents the satellite phase angle measured from its ascending node. The phase difference between satellites in adjacent planes is denoted by Δf. Similarly, the phase difference between two adjacent satellites is denoted by ΔΦ. ΔΩ represents the difference in right ascension (RAAN) of the ascending nodes between adjacent planes. ζ(u) represents the difference in longitude from the ascending node, and has two different values ​​depending on whether the satellite is in the ascending or descending phase. ζ(u) can be given by the following formula:

[0187]

[0188] Similarly, the satellite phase angle u can be determined as follows:

[0189]

[0190] Therefore, at a given time t, the latitude of the SSP location The longitude λ can be calculated as follows:

[0191]

[0192]

[0193] "Hop" indicates the relay process of data packets between two satellites. v and H h These are the hop counts for satellites in the same plane and those in different planes, respectively. Therefore, the phase difference Δu between the two satellites is given by the following formula:

[0194] Δu=u GS -u AS =H h Δf+H v ΔΦ (18)

[0195] Number of jumps in different planes H h The difference ΔL0 between the two plane satellites can be directly calculated. L0∈[-π,π] is the initial longitude of the ascending node. ΔL0 determines the relative position of the orbital plane, which can be obtained through equation (17). ΔL0 and H h It is given by the following formula:

[0196] ΔL0=Δλ+ζ(u AS )-ζ(u GS (19)

[0197]

[0198] The Round(x) function represents taking the integer closest to x.

[0199] Based on formula (18), the number of jumps in the same orbital plane can be expressed as:

[0200]

[0201] Finally, the total number of hops between AS and GS can be given by the following formula:

[0202] H = |H v |+|H h | (22)

[0203] Since AS and GS can be increasing or decreasing, equation (22) based on equations (14)(15)(20)(21) has four possible outcomes. Therefore, H can be determined by one of the following options: H A-D H D-D H A-A or H D-A H A-D This represents the number of hops from the ascending orbit access satellite to the descending orbit gateway satellite.

[0204] Therefore, the expression for the minimum hop count from the user to all gateway stations is:

[0205]

[0206] Where i represents the gateway station number, there are a total of M gateway stations, and x, y represent the satellite type (ascending orbit A or descending orbit D). This represents the hop count from an access satellite of type x to the corresponding satellite of type y at the i-th gateway station. Therefore, the access satellite type should be selected as follows:

[0207]

[0208] Where AS=1 represents selecting an ascending satellite as the access satellite type, and AS=0 represents selecting a descending satellite as the access satellite type.

[0209] Based on this evaluation method, users can quickly assess the type of satellite they should access using their own latitude and longitude, and always switch to the same type when switching satellites, thus ensuring that the satellites always maintain a low average latency and jitter.

[0210] Phase II, Multi-hop Aware Routing Phase:

[0211] In the second phase of routing, due to the limited location of the gateway station, data packets sent by users far from the gateway station may need to be transmitted over long distances. Furthermore, due to the uneven distribution of global traffic, the routing in this phase must ensure both low propagation latency and a certain level of congestion avoidance capability. In low-Earth orbit (LEO) mega-constellation networks, while distributed routing algorithms can guarantee real-time route calculation, they rely solely on the states of neighboring nodes, making them prone to getting trapped in local optima. Centralized algorithms, on the other hand, often require significant signaling overhead and cannot guarantee real-time performance for LEO mega-constellation systems. Therefore, this invention designs a semi-distributed routing algorithm for this routing phase, combining the advantages of both distributed and centralized algorithms while ensuring low signaling overhead. This algorithm enables fast routing with low latency and congestion avoidance capabilities.

[0212] 2.1 Multi-hop load awareness

[0213] Distributed algorithms typically make routing decisions based on the states of neighboring nodes. However, the half-shared approach differs from distributed algorithms, requiring more satellite node information to participate in the decision-making process. Therefore, a multi-hop load awareness zone is established for each satellite. To achieve load awareness within an n (n≥2) hop range, each satellite node periodically collects satellite information within an n-hop range from itself. This information primarily includes the satellite ID and its own link cost metric. However, if the congestion state of a satellite node changes beyond a certain threshold θ... c When this occurs, it is defined as a congestion state, and a limited flooding is immediately performed to the surrounding n-hop range. If each satellite establishes 4 inter-satellite links, then the total number of surrounding satellites that need to be collected within the n-hop range is 2n(n+1), such as... Figure 6 As shown, when n=3, the satellites within the purple rhombus area are the current satellite S. c =S ij The system senses a 3-hop sensing area containing 24 satellites. Red nodes represent congested nodes, green nodes represent normal nodes, and blue nodes represent gateway stations. Assuming... The satellite will then propagate data packets to the GS1 gateway station with fewer hops from the current satellite. Satellite nodes can construct a directed graph based on the collected information about surrounding satellites. The nodes of the graph are satellites, and the edges are inter-satellite links, distinguishing between output and input links. The weight of each edge is a link cost metric. Next, the optimal path will be found within the sensing area based on the constructed directed graph model.

[0214] 2.2 Semi-distributed routing strategy

[0215] Once the multi-hop load sensing area of ​​the current satellite is determined, routing decisions can be made within that area. When the access satellite S is determined according to formulas (23) and (24)... i,j After the gateway station GS1, from the access satellite S i,j Starting with satellite S i,j An n-hop load sensing area is established around the center. Priority is given to estimating the hop count from the satellite nodes at the edge of the sensing area (i.e., the nth hop node) to the gateway station GS1. There are 4n edge satellite nodes at the edge of the n-hop sensing range. Figure 6 In the middle, the set of edge nodes is {S} i,j+3 ,S i+1,j+2 ,S i+2,j+1 ,S i+3,j A total of 12 satellite nodes are selected. The edge node with the fewest hops and not in a congested state is selected as the target node within the sensing area (excluding congested nodes S). i+3,j If multiple n-hop nodes have the same minimum hop count, as shown in Figure S... i+1,j+2 ,S i+2,j+1 , respectively with S i+1,j+2 ,S i+2,j+1 Using GS1 as a vertex, two rectangular PathCandidate Regions (PCRs) can be drawn. Each PCR is a set of paths with a minimum hop count, such as... Figure 6 (The diagonal and horizontal lines in the text).

[0216] Assume the current satellite is S c1,c2 The target satellite is S d1,d2 Then from S c1,c2 To S d1,d2 The number of minimum hop paths N sp (H v H h )for:

[0217]

[0218] Among them, H v and H h These represent the minimum forwarding hops in the vertical and horizontal directions, respectively. In the Walker-Delta constellation, H... v =min{|d1-c1|,M p -|d1-c1|}andH h =min{|d2-c2|,N p -|d2-c2|}. When 2≤m≤n, the possible remaining hops in each direction after a data packet is forwarded once are H. v =m-2,H n = n-1 or H v =m-1,H h = n-2, according to (25) we can get:

[0219] N sp (m-1,n-2)>N sp (m-2,n-1) (26)

[0220] Formula (26) indicates that if a data packet is forwarded through a remaining path selection domain that is closer to the square, the number of feasible paths in the path selection domain after forwarding is relatively larger. The probability that a data packet can avoid congestion queuing from the current node to the destination node is defined as the path survival probability P. PCR (S c1,c2 ,S d1,d2 It can be proven that, under specific conditions, when the link congestion transition probability in the path selection domain is high, forwarding the next hop along the direction with more remaining hops maximizes the survival probability of subsequent paths. Figure 6 In, that is, P PCR (S i+1,j+2 ,GS1)>P PCR (S i+2,j+1 Therefore, prioritizing the selection of the remaining path region (i.e., the diagonal region) that approximates a square can improve congestion avoidance performance, and the optimal target node is node S. i+2,j+1 Next, we will use satellite S... i,j Starting from the node, with satellite S i+2,j+1 For the target node, use Dijkstra's algorithm to calculate satellite S. i,j To satellite S i+2,j+1 The shortest path is the routing strategy within the sensing area. Since the edge weights around congested satellites are very high, this strategy bypasses high-weight paths to avoid congestion. It's important to note that once the routing strategy is determined, data packets are forwarded directly within the sensing area according to the predetermined strategy until they reach an edge node, where a new routing decision is made. When a data packet is routed to edge node S...i+2,j+1 Afterwards, a new round of multi-hop sensing will begin, continuing with satellite S... i+2,j+1 The node constructs a sensing area within an n-hop range to make routing decisions until the data packet is routed to the multi-hop sensing area centered on the gateway station, and then enters the next routing stage.

[0221] Phase III, Load Balancing Phase:

[0222] Gateway stations are critical infrastructure connecting satellites and terrestrial networks. Gateway satellites are responsible for collecting data packets from surrounding satellites and are key nodes in the entire constellation. However, the data density in the sensing area is extremely high, and achieving load balancing within the sensing area helps increase the overall network throughput. This phase first establishes a multi-hop sensing area centered on the gateway station. When data packets are routed to a backpressure routing strategy based on link metrics, the throughput and delivery rate of the gateway station are improved.

[0223] 3.1 Multi-hop sensing area centered on gateway stations

[0224] Unlike the sensing area established in Section 2.1 above, the multi-hop sensing area centered on the gateway station is relatively fixed. It has been previously assumed that the gateway station can only access two satellites simultaneously at a constellation altitude: one ascending gateway satellite and one descending gateway satellite. Sensing areas with a range of m (m≥3) hops are established centered on these two gateway satellites. These sensing areas cover all ascending and descending satellites surrounding the gateway station, ensuring that data packets entering the sensing area are promptly transmitted to the gateway station.

[0225] 3.2 Backpressure routing strategy centered on gateway station

[0226] When a data packet is routed to the edge of the sensing area as described in the previous section, the switch to Phase III strategy begins. Since the link metric was calculated in the previous section, a backpressure routing strategy based on the link metric is designed next. Backpressure routing is an algorithm proposed in existing technology that dynamically allocates traffic on a multi-hop network based on the congestion gradient between neighboring nodes. Traditional backpressure routing does not construct a specific source-destination path, but only calculates the queue backlog for each time slot. The data transmission path is determined according to the queue backlog maximization criterion. Given G = (V, E), Indicates satellite node S a S b The ISL between [1,N], where a,b∈[1,N] P M P At time slot t, This indicates the backlog of the queue at satellite a. This represents the queue backlog at satellite b. Therefore, the difference in queue backlog between a and b can be expressed as: Assuming that the same packet stream typically converges on a single satellite, therefore The backlog difference of the largest queue can be defined as

[0227]

[0228] In the link The following strategy is used to allocate transmission rates to the data stream:

[0229]

[0230]

[0231] in For link The transmission rate, Γ s(t) It contains a matrix of all possible transmission rates and can be scheduled according to network topology.

[0232] The basic theory of backpressure routing described above is based on the assumption that each satellite node has only one common queue. Since the satellites in this invention are designed with one queue per port (i.e., each satellite has four queues), the congestion state of the satellites can be described more accurately. Therefore, we define a new backpressure routing metric, defining the backlog metric in the direction of a certain port of satellite a as follows: x∈{1,2,3,4}, where x represents the four port directions.

[0233] Link at time slot t The backlog metric is:

[0234]

[0235] Wherein, the x-port direction of satellite a is the port connecting to satellite b, and the y-port direction of satellite b is the next-hop port under the minimum hop count path area constraint formed by satellite b and the target satellite. Then, in the link The above assigns a transmission rate to the data stream as a solution to the maximum weight algorithm based on link cost, as shown in Equation 32:

[0236]

[0237] in, It satisfies formula (29). During the routing process, when a transmission is scheduled, a data packet is dequeued at the head of the sending buffer of node a. If this data packet is selected on the link... For transmission in the middle, b must satisfy:

[0238]

[0239] Where N *This refers to all neighboring nodes of satellite a within the minimum hop count path region.

[0240] by Figure 7 For example, within the minimum hop count path range formed by the current node and the gateway satellite node, find the next hop node with the largest backlog difference, i.e.

[0241]

[0242]

[0243]

[0244] Among them, S next For S a Next-hop satellite node. The next-hop satellite node can be determined using formulas (34)-(36) until the data packet is transmitted to the gateway station.

[0245] In summary, the overall process of the hybrid routing method under the condition of limited gateway station deployment in the low-Earth orbit mega-constellation network provided in this embodiment is as follows: Figure 8 As shown.

[0246] The hybrid routing algorithm proposed in this invention is particularly suitable for specific mega-constellation scenarios where gateway stations are deployed in confined areas and data packets need to be transmitted over long distances to ground stations. This algorithm exhibits significant congestion avoidance and load balancing capabilities, especially under conditions of global load imbalance and large data volumes around gateway satellites or gateway stations.

[0247] It is worth noting that the various sub-stages in the algorithm are adaptable to diverse application scenarios. For example, the first stage, as an access strategy, is compatible with most routing algorithms, while the second stage is suitable for communication scenarios with globally distributed connections between any two ground users. Similarly, the third stage extends its applicability to scenarios with globally distributed gateway stations. Furthermore, all stages are modular, allowing for flexible combination of parts of the stages to reduce overhead.

[0248] The main advantage of the algorithm proposed in this invention lies in its successful integration of the advantages of centralized and distributed algorithms. This hybrid approach maintains a global perspective while preserving rapid response capabilities, combining the strengths of various sub-strategies to enhance network load performance and quality of service. The sensing area is highly adaptable and can be adjusted according to constellation scale.

[0249] Secondly, embodiments of the present invention provide a hybrid routing system under conditions of limited gateway station deployment in a low-Earth orbit mega-constellation network, as shown in the attached figure. Figure 9 As shown, it includes:

[0250] The access selection module is configured to calculate the minimum hop count from the access satellite to each gateway station based on the geographical location of the ground user and the hop count evaluation model, and select the combination of access satellite and gateway station with the minimum hop count.

[0251] The multi-hop sensing routing module is configured to establish a multi-hop load sensing area in the transmission path from the access satellite to the gateway station, dynamically plan the path according to the link cost metric model, and calculate the link cost by weighting queue occupancy rate, congestion probability and bandwidth utilization, and bypass congested nodes.

[0252] The load balancing module is configured to distribute traffic within the coverage area of ​​the gateway station using an improved backpressure routing strategy based on the queue backlog difference of the satellite ports, thereby achieving load balancing.

[0253] The present invention provides a hybrid routing system under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network, which is used to implement the hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network as described above. Its working principle will not be described in detail here.

[0254] The following specific experiments illustrate the beneficial effects of the hybrid routing method provided in the embodiments of the present invention.

[0255] The performance of the routing algorithm was evaluated using the OMNeT++ simulation platform. The simulation adopted a Walker-Delta constellation configuration, including 24 orbital planes, 22 satellites per orbit, a phase factor of 11, an altitude of 550 km, and an orbital inclination of 53°. Each satellite has two intra-orbit inter-satellite links and two inter-orbit inter-satellite links. The simulation duration was 60 minutes. All uplink, downlink, and inter-satellite link bandwidths were 10 Gbps. It was assumed that only three gateway stations were deployed in Sanya, Harbin, and Urumqi, while eight data streams distributed across continents around the world sent data packets to the gateway stations, always sending a constant data stream. Each data packet was 1024 B in size, and background traffic was set based on global population distribution.

[0256] The comparison methods employ distance-based backpressure routing (DBPR), Dijkstra's shortest path (DSP) based on the Dijkstra algorithm, and geolocation-based greedy perimeter stateless routing (GPSR). DSP is a centralized routing strategy, while DBPR and GPSR are distributed routing strategies. DBPR, being a distance-based backpressure routing strategy, possesses excellent load balancing capabilities. In GPSR, we made improvements by using a weighted sum of satellite load and distance as the evaluation metric to enhance load balancing. The following metrics are used to measure network performance:

[0257] 1) Delivery rate: The ratio of the total number of data packets received by each gateway station to the total number of data packets sent by the source node;

[0258] 2) Throughput: The average rate at which each gateway successfully receives data packets;

[0259] 3) Average packet hops: The average number of hops required for a gateway station to receive a data packet;

[0260] 4) Average delay: The average time required for all data packets to be sent from the source node to the gateway.

[0261] like Figure 10 As shown, when the Constant Data Rate (CBR) increases, packet loss occurs due to increased satellite load, and the delivery rate of all algorithms decreases. Among all algorithms, the proposed algorithm consistently maintains the highest delivery rate, even under conditions of high network congestion. This is achieved through the use of Phase II and Phase III strategies, ensuring network traffic balance. When the CBR rate reaches 10 Mbps, the proposed algorithm (HRSCM) achieves a data transmission rate nearly 27.1% higher than DSP. Furthermore, because the DBPR algorithm uses a distributed backpressure routing strategy, it employs the approximate delay between intermediate and destination satellites as link weights. This allows for flexible utilization of differential backpressure between neighboring nodes based on destination distance delay to balance congestion and distance, resulting in a high delivery rate.

[0262] like Figure 11 As shown, the throughput of all algorithms increases with increased traffic injection. It can be seen that the proposed algorithm achieves the maximum throughput, while the DSP algorithm achieves the minimum. DBPR, based on backpressure routing, is second only to the proposed algorithm. Although the GPSR algorithm is a distributed algorithm, its performance is lower than both DBPR and the algorithm of this invention. The proposed algorithm (HRSCM) consistently maintains high throughput because it uses a Phase III strategy to perform traffic load balancing near the gateway station. It can be seen that under a traffic rate of 10Gbps, throughput can be improved by 5.5%-62.1%.

[0263] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Additionally, the terms "front," "back," "left," "right," "upper," and "lower" in this document refer to the placement shown in the accompanying drawings.

[0264] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A hybrid routing method under conditions of limited gateway station deployment in a low-Earth orbit mega-constellation network, characterized in that, include: Access satellite and gateway station selection: Based on the geographical location of the ground user, the minimum number of hops from the access satellite to each gateway station is calculated using a hop count evaluation model, and the combination of access satellite and gateway station with the minimum number of hops is selected; Multi-hop aware routing decision: A multi-hop load awareness area is established in the transmission path from the access satellite to the gateway station. The path is dynamically planned according to the link cost measurement model. The link cost is calculated by weighting queue occupancy rate, congestion probability and bandwidth utilization, and the congested nodes are bypassed. Gateway station load balancing control: Within the coverage area of ​​the gateway station, traffic is distributed using a backpressure routing strategy based on the queue backlog difference of satellite ports to achieve load balancing.

2. The hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network according to claim 1, characterized in that, The hop count-based evaluation model includes: Based on the latitude and longitude of the ground user and the ascent or descent type of the access satellite, the minimum hop count from the access satellite to each gateway station is calculated. The combination of the access satellite and gateway station with the minimum hop count is selected, where the hop count is determined by a weighted sum of the hop counts in the same orbital plane and the hop counts in different orbital planes.

3. The hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network according to claim 1, characterized in that, The link cost metric is calculated based on the following parameters: The queue occupancy rate of a satellite node indicates the current buffer status of the satellite output port queue; Queue congestion probability is used to predict the congestion status of satellite nodes by combining the base probability and trend factor through a sliding time window. Link bandwidth utilization is dynamically assessed using a sliding time window and exponential moving average method to evaluate the real-time load of the link. The link cost metric is obtained by weighted summation and applying a nonlinear penalty term, which causes the cost of high-load links to increase rapidly.

4. The hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network according to claim 3, characterized in that, The calculation of the queue congestion probability includes the following steps: Construct a fixed-size sliding time window to record the queue occupancy rate sequence at several recent time points; Calculate the base probability, which is determined based on the ratio of the current queue occupancy rate to a preset congestion threshold; A trend factor is calculated by comparing the difference between the current queue occupancy rate and the average occupancy rate within the sliding time window, and then normalizing the difference. The queue congestion probability is calculated by combining the basic probability and trend factor to predict the congestion status of satellite nodes.

5. The hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network according to claim 3, characterized in that, The calculation of the link bandwidth utilization includes the following steps: Construct a fixed-size sliding time window, periodically sample the data transmitted through the link, and record the cumulative number of bytes transmitted at each sampling point; Calculate the instantaneous bandwidth utilization rate, which is the ratio of the actual number of bytes transmitted within the window to the theoretical maximum transmission capacity; The instantaneous bandwidth utilization is smoothed using the exponential moving average method to obtain the current bandwidth utilization. The smoothing factor is between 0 and 1 and is used to balance instantaneous responsiveness and stability.

6. The hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network according to claim 1, characterized in that, The construction of the multi-hop load sensing region includes: Centered on the current satellite, periodically collect the congestion status and link cost of satellites within a range of n hops, where n≥2; When the queue occupancy rate of a satellite node exceeds a threshold, a limited flooding update of the neighbor node status is triggered. Within the sensing area, the Dijkstra algorithm is used to calculate the shortest path, prioritizing the edge nodes with the fewest hops and the lowest link cost.

7. The hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network according to claim 6, characterized in that, When multiple edge nodes have the same number of hops, a remaining path selection domain that is close to a square is selected to improve the path survival probability, where the path survival probability is defined as the probability of avoiding congested links.

8. The hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network according to claim 1, characterized in that, The backpressure routing strategy includes: An n-hop sensing region is constructed with the gateway satellite as the center, covering the ascending and descending satellites around the gateway station, where n≥2; Calculate the backlog difference between adjacent satellites based on the queue backlog metric of each satellite port; Within the path area with the minimum number of hops, the next hop node with the largest backlog difference is selected for data packet transmission until it reaches the gateway station.

9. The hybrid routing method under the condition of limited gateway station deployment in a low-Earth orbit mega-constellation network according to claim 8, characterized in that, The queue backlog metric is based on the independent queues of each output port of the satellite, defined as the backlog in a certain port direction, and optimizes the transmission rate allocation through the link cost metric.

10. A hybrid routing system under conditions of limited gateway station deployment in a low-Earth orbit mega-constellation network, characterized in that, include: The access selection module is configured to calculate the minimum number of hops from the access satellite to each gateway station based on the geographical location of the ground user and the hop count evaluation model, and select the combination of access satellite and gateway station with the minimum number of hops. The multi-hop sensing routing module is configured to establish a multi-hop load sensing area in the transmission path from the access satellite to the gateway station, dynamically plan the path according to the link cost metric model, wherein the link cost is calculated by weighting queue occupancy rate, congestion probability and bandwidth utilization, and bypass congested nodes. The load balancing module is configured to distribute traffic within the coverage area of ​​the gateway station using an improved backpressure routing strategy based on the queue backlog difference of the satellite ports, thereby achieving load balancing.

Images