A time-varying routing method and apparatus for satellite networks

Abstract

A time-varying routing method for satellite networks, applied to satellite network routing, includes: dividing the system cycle into atomic snapshot sequences; performing optimal segmentation of the atomic snapshot sequences using a state-extended dynamic programming algorithm to generate a global snapshot scheduling table containing aggregate snapshot sequences and their corresponding start and end times; distributing the global snapshot scheduling table to each satellite node and loading it; constructing and maintaining a network-wide logical topology mask based on the logical topology defined by the current snapshot in the global snapshot scheduling table, wherein the logical topology mask records the logical state of all inter-satellite links in the network; collaboratively maintaining the logical topology mask through time-triggered and event-triggered mechanisms; and calculating a routing table based on the links in the logical topology mask whose link state is connected and distributing it to the data plane. This method enables the time-varying routing mechanism to balance the routing update overhead of the control plane and the link utilization of the data plane when dividing snapshots.

Description

Technical Field

[0001] This invention relates to the field of routing technology, and in particular to a time-varying routing method and apparatus for satellite networks. Background Technology

[0002] Low Earth Orbit (LEO) satellite networks, with their significant advantages of low latency and global coverage, have become an important component of the next-generation internet infrastructure. Currently, large-scale LEO satellite constellations, represented by Starlink, Kuiper, and GW constellations, are under large-scale construction. Unlike terrestrial networks with relatively fixed topologies, LEO satellites move at high speeds relative to the ground, resulting in highly dynamic and time-varying network topologies: In high-latitude regions, the relative angular velocity between adjacent orbital satellites increases dramatically, making it difficult for lasers at both ends of the link to align, leading to the closure of inter-satellite links between different orbits; in regions with alternating day and night, when the angle between the pointing axis of the laser transmitter and sunlight is less than the safe angle, satellite nodes will shut down their laser transmitters to avoid sunlight and prevent permanent thermal damage to their optical systems, causing inter-satellite link interruptions; furthermore, topology change events are not uniformly distributed over time, with link interruptions in high-latitude regions showing a concentrated burst over time, while inter-satellite link interruptions caused by sunlight avoidance are scattered over time. The highly dynamic nature of satellite network topology poses a severe challenge to routing protocols: if traditional routing protocols are used directly, frequent topology changes will trigger a flood of link state updates across the entire network, causing signaling storms across the network, resulting in routing oscillations and severely damaging network stability. Summary of the Invention

[0003] To address the problems existing in the prior art, embodiments of this application provide a method, apparatus, computing device, computer storage medium, and product containing a computer program for time-varying routing in satellite networks. This enables the time-varying routing mechanism to balance the routing update overhead of the control plane and the link utilization of the data plane when dividing snapshots, thereby improving network stability.

[0004] In a first aspect, embodiments of this application provide a time-varying routing method for satellite networks, applied to routing in satellite networks. The method includes: dividing the system period into atomic snapshot sequences; performing optimal segmentation of the atomic snapshot sequences using a state-extended dynamic programming algorithm to generate a global snapshot scheduling table containing aggregate snapshot sequences and their corresponding start and end times; distributing the global snapshot scheduling table to each satellite node and loading it; constructing and maintaining a network-wide logical topology mask based on the logical topology defined by the current snapshot in the global snapshot scheduling table, wherein the logical topology mask records the logical state of all inter-satellite links in the network; collaboratively maintaining the logical topology mask through a time-triggered mechanism and an event-triggered mechanism; and calculating a routing table based on the links in the logical topology mask whose link state is connected and sending it to the data plane.

[0005] In some possible implementations, the system cycle is divided into a sequence of atomic snapshots. A state-extended dynamic programming algorithm is used to optimally segment this sequence, generating a global snapshot scheduling table containing aggregated snapshot sequences and their corresponding start and end times. This includes: dividing the system cycle into M atomic snapshots of equal length to form an atomic snapshot sequence; and using the index space of the atomic snapshot sequence... In the given information, determine an ordered set of interval coordinate pairs. The atomic snapshots within each interval are merged into an aggregate snapshot to generate an aggregate snapshot set. The logical topology corresponding to each element in the aggregate snapshot set is formed by the intersection of the topologies of all atomic snapshots within the interval. The optimal interval sequence set and the number of segments are solved by using the state extension dynamic programming algorithm with the objective function of minimizing the total cost of the entire system.

[0006] In some possible implementations, the objective function is:

[0007]

[0008]

[0009] In the formula, It is an interval The total cost of merging atomic snapshots into a single aggregate snapshot consists of two parts: routing update cost and performance distortion cost.

[0010] In some possible implementations, the cost of route updates includes computational overhead. Storage overhead Path switching overhead The performance distortion comes at the cost of wasted link capacity. and delay penalty The formula for calculating the total cost is as follows:

[0011]

[0012] In the formula, It is a first-level weighting factor and , It is a second-order weighting factor and satisfies , .

[0013] In some possible implementations, path switching overhead is defined as the cumulative sum of the absolute changes in the shortest path latency of all node pairs across the network when switching from the previous aggregate snapshot to the current aggregate snapshot.

[0014] Among some possible implementations, the state-extended dynamic programming algorithm includes: definition This indicates the atomic snapshot range. The data is divided into segments, and the last aggregated snapshot is the merged interval. Given an atomic snapshot, find the global minimum normalization total cost; establish the transition state equation:

[0015]

[0016] In the formula, is In-snapshot cost, To merge intervals Switch the aggregated snapshot to the merged range The path switching overhead of the aggregated snapshot is calculated; the ending state that minimizes the total cost of the entire cycle is obtained by recursion, and the optimal interval sequence set is reconstructed using the predecessor of the record.

[0017] In some possible implementations, generating a global snapshot schedule table containing the aggregated snapshot sequence and its corresponding start and end times also includes: limiting the maximum time slice span of the aggregated snapshots. The search space of dynamic programming is constrained within a sliding window; during the process of traversing and building aggregate snapshots, the overall network connectivity of the intersection topology is detected in real time, and when a physical disconnection is detected, the subsequent expansion traversal under the current starting point is immediately stopped.

[0018] In some possible implementations, the time-triggered mechanism includes: when the system clock reaches the snapshot switching time, the satellite node reads the topology definition of the next snapshot in the global snapshot scheduling table; queries the status of all links in the logical topology mask; if the link status is faulty, the fault status remains unchanged; if the link status is shielded or connected, the link status of the logical topology mask is set to be consistent with the snapshot topology.

[0019] In some possible implementations, the event triggering mechanism includes local event triggering and remote event triggering: Local event triggering refers to querying the status of the corresponding link in the logical topology mask when the status of the physical link directly connected to the local node changes. If it is in a masked state, the change is masked and the generation of link status update messages is suppressed; if it is in a connected or faulted state, the logical topology mask is updated and a link status update message is generated and flooded to the entire network. Remote event triggering refers to querying the status of the corresponding link in the logical topology mask when a link status update message flooded by a neighbor is received. If it is in a masked state, the message is masked; if it is in a connected or faulted state, the logical topology mask is updated and the message is flooded to other neighbors.

[0020] Secondly, embodiments of this application provide a time-varying routing device for a satellite network, comprising: a snapshot partitioning module, used to divide the system period into atomic snapshot sequences, perform optimal partitioning of the atomic snapshot sequences using a state-extended dynamic programming algorithm, and generate a global snapshot scheduling table containing aggregate snapshot sequences and their corresponding start and end times; a distribution and loading module, used to distribute the global snapshot scheduling table to each satellite node and load it; a logical topology mask construction module, used to construct and maintain a network-wide logical topology mask based on the logical topology defined by the current snapshot in the global snapshot scheduling table, wherein the logical topology mask records the logical state of all inter-satellite links in the network; a maintenance module, used to maintain the logical topology mask collaboratively through a time-triggered mechanism and an event-triggered mechanism; and a routing calculation module, used to calculate a routing table based on the links in the logical topology mask whose link state is connected and distribute it to the data plane.

[0021] Thirdly, embodiments of this application provide a computer-readable storage medium including computer-readable instructions that, when read and executed by a computer, cause the computer to perform the method as described in any of the first aspects.

[0022] Fourthly, embodiments of this application provide a computing device, including a processor and a memory, wherein the memory stores computer program instructions, which, when executed by the processor, perform the method as described in any of the first aspects.

[0023] Fifthly, embodiments of this application provide a product comprising a computer program that, when the computer program product is run on a processor, causes the processor to perform the method as described in any of the first aspects. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0025] Figure 1 This is a schematic diagram of a low-Earth orbit satellite network system provided in an embodiment of this application;

[0026] Figure 2 This is a schematic diagram of a time slice structure provided in an embodiment of this application;

[0027] Figure 3 This is a flowchart illustrating a time-varying routing method for satellite networks provided in an embodiment of this application;

[0028] Figure 4This is a schematic diagram of snapshot partitioning provided in an embodiment of this application;

[0029] Figure 5 This is a schematic diagram of a dynamic programming state transition provided in an embodiment of this application;

[0030] Figure 6 This is a schematic diagram illustrating the generation, distribution, and loading process of a global snapshot scheduling table provided in an embodiment of this application;

[0031] Figure 7 This is a schematic diagram illustrating three types of link state machine transitions provided in an embodiment of this application;

[0032] Figure 8 This is a schematic diagram of a time-triggered process provided in an embodiment of this application;

[0033] Figure 9 This is a schematic diagram of a local event triggering process provided in an embodiment of this application;

[0034] Figure 10 This is a schematic diagram of a remote event triggering process provided in an embodiment of this application;

[0035] Figure 11 This is a schematic diagram illustrating the calculation performed by a routing execution device according to an embodiment of this application;

[0036] Figure 12 This is a schematic diagram of the structure of a time-varying routing device for a satellite network provided in an embodiment of this application. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] In this article, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The symbol " / " in this article indicates that the related objects are in an "or" relationship; for example, A / B means A or B.

[0039] The terms "first" and "second," etc., used in the specification and claims herein are used to distinguish different objects, not to describe a specific order of objects. For example, "first response message" and "second response message," etc., are used to distinguish different response messages, not to describe a specific order of response messages.

[0040] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0041] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.

[0042] To facilitate understanding of the embodiments of this application, the following will provide further explanation and description with reference to the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0043] Figure 1 A schematic diagram of a low-Earth orbit satellite network system is shown. Figure 1 As shown, a low-Earth orbit (LEO) satellite network system comprises a space segment, a ground segment, and a user segment. The user segment is the terminal access portion of the LEO satellite network, covering various users and terminal devices that require communication and networking via the LEO satellite network. Various terminals in the user segment establish connections with LEO satellites through satellite-to-ground links, accessing the satellite network and enabling data transmission and reception. The ground segment connects satellites to the ground via ground stations. The space segment consists of a large-scale LEO satellite constellation and is crucial for achieving data relay in the air and global coverage. Satellite nodes in mainstream constellations are equipped with laser inter-satellite link devices, establishing interconnections between satellites to form a space backbone transmission network, enabling high-speed data relay between satellite nodes. Simultaneously, routing technology is the core support at this level, responsible for coordinating the interconnection rules and data forwarding paths between satellites, ensuring efficient data transmission.

[0044] The entire low-Earth orbit satellite network moves at high speed relative to the ground. This highly dynamic characteristic poses a severe challenge to routing protocols: if traditional routing protocols are used directly, frequent topology changes will trigger a flood of link-state updates across the entire network, causing signaling storms, routing oscillations, and severely damaging network stability. Currently, the mainstream routing scheme for the highly dynamic characteristics of satellite networks is a snapshot-based time-varying routing mechanism. For example, Figure 2A schematic diagram of a time-slice structure is shown. The snapshot-based time-varying routing mechanism discretizes the continuous time axis into a series of static time slices, and uses ephemeris data (a set of mathematical parameters describing the characteristics of satellite orbits, used to accurately calculate and predict the three-dimensional spatial coordinates of satellites at any time) to predict the continuously connected links in each time slice, and calculates the topological snapshot of each time slice, thereby transforming the dynamic time-varying graph into a series of static graph sequences.

[0045] While snapshot-based time-varying routing mechanisms have addressed the routing oscillation problem caused by dynamic nature to some extent, existing snapshot partitioning strategies struggle to balance routing stability and link utilization, and are prone to snapshot fragmentation. The discretization of time-varying topology leads to a loss of link capacity and consequently, transmission performance. Traditional equal-length snapshot partitioning cannot adapt to the non-uniform distribution of topology change events over time. During periods of dense topology change, it causes topology distortion and wasted link capacity; during periods of sparse topology change, it leads to unnecessary snapshot switching. While existing adaptive snapshot strategies introduce variable-length time slices, they are mostly performance-oriented, aiming to minimize the number of snapshots while ensuring path delay increments or network capacity, often ignoring the routing oscillation costs caused by path changes during snapshot switching. During periods of dense topology change, this strategy tends to fragment snapshots extremely finely to maintain theoretically optimal transmission performance, resulting in frequent snapshot switching. This excessive fragmentation has serious side effects: on the one hand, it leads to frequent route calculations and updates, consuming the computing resources of onboard routers; on the other hand, frequent path switching can cause packet out-of-order delivery, rendering the TCP congestion window ineffective and ultimately reducing the actual network throughput.

[0046] Furthermore, current methods cannot distinguish between predictable and unpredictable topology changes, leaving the network still vulnerable to signaling storms. Traditional snapshot-based time-varying routing mechanisms pre-calculate full-cycle snapshot topologies offline based on ephemeris data and update routes accordingly. However, this mechanism lacks awareness of unpredictable topology changes occurring during network operation. Once such changes occur, the pre-built static routing table will partially fail, leading to routing interruptions and packet loss. To address fault tolerance issues, existing improvements attempt to introduce a hybrid approach combining dynamic routing protocols with snapshot-based time-varying routing mechanisms. This hybrid routing mechanism can handle both predictable and unpredictable topology changes. However, this simple hybrid routing mechanism cannot distinguish between predictable and unpredictable topology changes, resulting in unnecessary link-state update flooding even when predictable topology changes occur, exposing the entire network to signaling storm risks.

[0047] In view of this, this application provides a time-varying routing method for satellite networks. Addressing the problem that existing snapshot partitioning strategies struggle to balance routing stability and link utilization, and are prone to snapshot fragmentation leading to frequent route updates and TCP congestion window failures, a snapshot partitioning method for low-Earth orbit satellite networks based on state-extended dynamic programming is proposed. First, snapshot partitioning is modeled as an optimal segmentation problem for a time series. Aggregate snapshots are constructed by determining a set of ordered interval coordinate tuples in the atomic snapshot sequence. Within an interval, the intersection of all atomic snapshot topologies is used as the logical topology for the corresponding time period. To overcome the shortcomings of traditional algorithms that only focus on performance indicators while ignoring snapshot switching costs, this embodiment constructs a comprehensive cost model that includes path switching overhead. This model defines path switching overhead as the cumulative sum of the absolute changes in the shortest path delay for all nodes in the entire network when adjacent aggregate snapshots switch, used to quantify the impact of topology changes on network stability. The total cost function of aggregate snapshots is composed of a weighted average of normalized route update costs and performance distortion costs. Furthermore, considering that path switching overhead is related to both the current snapshot and the previous snapshot, an extended dynamic programming algorithm in a two-dimensional state space is designed. By introducing the preceding snapshot index as an extended state variable, the evolutionary relationship between adjacent snapshots is accurately captured, and a global snapshot scheduling table covering the entire system cycle with the lowest overall cost is recursively solved. Furthermore, to reduce the computational complexity of the state-extended dynamic programming algorithm under high-granularity atomic snapshots, a dual acceleration strategy based on the physical characteristics of low-Earth orbit satellites is introduced: firstly, search space pruning is performed through physical constraints to limit the maximum time slice span of aggregated snapshots. First, the nested loops of dynamic programming are strictly constrained within a sliding window. Second, the monotonic sparsity of the topology is utilized to introduce an early termination mechanism for connectivity breaks. When a physical break occurs in the intersection topology within an aggregate snapshot, redundant topology expansion and routing calculations for that interval are immediately truncated. Through these optimizations, the complexity of the original algorithm is significantly reduced, greatly improving the solution efficiency of large-scale constellation offline snapshot partitioning algorithms in practical engineering. Furthermore, addressing the problem that existing time-varying routing mechanisms cannot distinguish between predictable and unpredictable topology changes, leading to signaling storms in the network, this application employs a dual-trigger distributed routing method based on a full-network logical topology mask. A full-network logical topology mask, independent of the underlying physical interface, is constructed on the satellite nodes. This mask maintains the state of all links in the network, including connected, masked, and faulty link states, serving as a dynamic benchmark for routing decisions. The mask is maintained using a time-triggered and event-triggered collaborative mechanism: at snapshot switching time, the mask state is reset according to the current snapshot's topology definition, and the shielding state of the links is used to suppress predictable topology changes that generate link state update messages. Simultaneously, the fault state of the links from the previous snapshot is automatically inherited, achieving zero-signaling snapshot switching. During the snapshot's existence, the physical link state and neighbor signaling are monitored in real time. Physical layer events are compared with the logical mask state. Changes in the state of links in the shielding state are judged as predictable topology changes, and signaling generation is suppressed. Sudden failures of links in the connected state and the recovery of links in the faulty state are judged as unpredictable events, and the mask is immediately updated, triggering a network-wide link state update and route recalculation. The method provided in this application enables the time-varying routing mechanism to balance the routing update overhead of the control plane and the link utilization of the data plane when dividing snapshots, and to distinguish between predictable and unpredictable topology changes in distributed satellite networks.

[0048] For example, Figure 3 A flowchart illustrating a time-varying routing method for satellite networks, provided in an embodiment of this application, is shown. This method is applied to routing within a satellite network. Figure 3 As shown, the method may include the following steps:

[0049] S31: The topology management device of the ground station divides the system cycle into atomic snapshot sequences, and performs optimal segmentation of the atomic snapshot sequences through a state extension dynamic programming algorithm to generate a global snapshot scheduling table containing aggregate snapshot sequences and their corresponding start and end times.

[0050] In this embodiment, to balance the routing update overhead of the control plane and the link utilization of the data plane, a snapshot partitioning of the low-Earth orbit satellite network based on state extension dynamic programming is first required. The mathematical model upon which the snapshot partitioning is based is defined as follows: the system period T of the satellite network is divided into M (M>1) equal-length, fine-grained time slices, forming an atomic snapshot sequence, denoted as... Each atom snapshot This represents a time interval The physical network topology state within, where τ is the atomic time granularity, and has Since τ is sufficiently small, we assume that the topology remains static within a single atomic snapshot. The snapshot partitioning problem is defined as follows: in the index space of a sequence of atomic snapshots... In the given condition, we determine an ordered set of N interval coordinate pairs, denoted as . The set must satisfy the following continuity and boundary constraints:

[0051] Intra-interval constraints: For any interval The starting point is no later than the ending point, that is .

[0052] Continuity constraint: for any adjacent intervals and ,have .

[0053] Boundary constraints: cover the entire cycle, i.e. and .

[0054] For example, Figure 4 This illustration shows a snapshot partitioning diagram provided by an embodiment of this application. Figure 4 As shown, determine the set of binary pairs. Next, the atomic snapshots within each interval are merged into an aggregate snapshot, generating the corresponding aggregate snapshot set. Aggregated snapshots Corresponding to the atomic snapshot interval That is, it covers the atomic snapshot sequence. In each aggregate snapshot During its lifetime, the network adopts a unified logical topology. Perform route calculation. The logical topology is formed by the intersection of all atomic snapshot topologies within the interval: .

[0055] The core of the snapshot partitioning problem lies in finding the optimal set of interval sequences. Given the number of segments K, the objective function for minimizing the total system cost over the entire cycle is expressed as:

[0056]

[0057]

[0058] In the formula, It is an interval The total cost of merging atomic snapshots into a single aggregate snapshot is comprised of the route update cost. With the cost of performance distortion It consists of two parts. Routing update cost This reflects the resource consumption and path switching oscillations in the control plane caused by route updates, including computational overhead. Storage overhead Path switching overhead Performance distortion cost This reflects the reduced data plane transmission capacity, including wasted link capacity, caused by the distortion of the merged aggregate snapshot compared to the atomic snapshot. and delay penalty Total cost The calculation formula is as follows:

[0059]

[0060] In the formula, It is a first-level weighting factor used to adjust the policy preference between the routing stability of the control plane and the transmission efficiency of the data plane. It is a secondary weighting factor used for fine-grained adjustment of the relative importance of each sub-indicator, and satisfies... , .

[0061] Furthermore, calculate the overhead. Defined as the computational resources consumed by the onboard router in generating a new routing table using the shortest path algorithm; storage overhead. The storage space occupied by the topology used to store aggregate snapshots; path switching overhead. Defined as the cumulative sum of the absolute changes in shortest path latency for all node pairs across the entire network when switching from the previous aggregate snapshot to the current aggregate snapshot; link capacity waste. Defined as idle link bandwidth resources that are physically connected but disconnected in the aggregate snapshot; latency penalty Defined as the latency increment of the shortest path in the aggregate snapshot compared to the shortest path in each atomic snapshot.

[0062] Furthermore, calculate the overhead. Storage overhead Link capacity waste Delay penalty Path switching overhead is only relevant to the properties of the current aggregate snapshot. This is related to the properties of both the current aggregate snapshot and the previous aggregate snapshot. Therefore, the cost within the snapshot is defined. Indicates merged intervals The cost of obtaining an aggregated snapshot from an atomic snapshot; defining the cost of switching between snapshots. This indicates a snapshot from the previous aggregated snapshot (which merged the intervals). Switch from the atomic snapshot to the current aggregate snapshot (which merges the intervals) The path switching overhead of (atomic snapshot), when At that time, there is no previous snapshot, let .

[0063] Since the cost of switching between snapshots depends not only on the attributes of the current aggregated snapshot but also on the attributes of the previous snapshot, this invention employs a state-extended dynamic programming algorithm to solve the snapshot partitioning problem. Please refer to [reference needed]. Figure 5 ,definition This indicates the atomic snapshot range. The data is divided into segments, and the last aggregated snapshot is the merged interval. Given an atomic snapshot, the global minimum normalization total cost is achieved. , In order to calculate It is necessary to enumerate the starting point of the previous snapshot. , find the state that makes Transition to state The path with the minimum cost has the following state transition equation:

[0064]

[0065] For the first segment of the sequence (i.e.) No snapshot switching is required. .

[0066] Generating a global snapshot schedule table containing aggregated snapshot sequences and their corresponding start and end times may specifically include the following steps:

[0067] S311: Topology preprocessing.

[0068] For each atomic snapshot starting point Let the termination point from Traverse to the physically constrained boundary. Constructing merged intervals The logical topology corresponding to the aggregated snapshot obtained from the atomic snapshot Because the intersection topology has a strictly monotonically sparse property when expanding along the time axis (i.e. ), during traversal During the process, the system monitors in real time. Network connectivity. If detected... If a physical disconnection occurs (i.e., an unreachable node pair exists), the early termination mechanism is triggered, and the current connection is immediately stopped. Next steps Extended traversal, and interval and all subsequent intervals In-snapshot cost Marked as infinity (∞). For maintaining connectivity Furthermore, its shortest path matrix across the entire network is calculated and cached for subsequent rapid calculation of latency parameters.

[0069] S312: Cost calculation within a snapshot.

[0070] For all merged intervals The aggregated snapshot obtained from the atomic snapshot is used to calculate its in-snapshot cost. This cost consists of computational overhead, storage overhead, wasted link capacity, and latency penalties. Among these, computational overhead... The calculation formula is:

[0071]

[0072] In the formula, It refers to the number of network nodes. The number of links is relatively fixed. It is the number of CPU instruction cycles required per unit of computational complexity.

[0073] Storage overhead The calculation formula is:

[0074]

[0075] In the formula, It is an interval The number of "stable links" that remain connected throughout the system. It represents the number of bytes used to store a single link descriptor.

[0076] Link capacity waste The calculation formula is:

[0077]

[0078] In the formula, It is the set of links in the k-th atomic snapshot. It is a collection of links in the aggregated snapshot. It is a link The bandwidth.

[0079] Delay penalty The calculation formula is:

[0080]

[0081] In the formula, This is the topological graph of the k-th atom snapshot. It is a topology graph of aggregated snapshots. This represents the shortest path delay from source node s to destination node d in graph G.

[0082] Since the various indicators have different physical dimensions, direct summation has no practical physical meaning; therefore, each indicator needs to be normalized. For computational and storage overhead, it is mapped to the proportion of limited onboard satellite resources used. Normalization of computational overhead... , It is the maximum number of instructions per second for the onboard processor. It is the time slice length allocated for route computation; normalized storage overhead. , This is the available storage space for the onboard router. For performance distortion, it is mapped to a performance degradation rate relative to an atomic snapshot. Normalized capacity waste. , It is an interval The sum of the physical bandwidths of all atomic snapshots within the range; normalized latency penalty , It is an interval The sum of the time delays between all atomic snapshot node pairs within the node.

[0083] Calculate merged intervals In-snapshot cost of the aggregated snapshot obtained from the atomic snapshot:

[0084]

[0085] The in-snapshot costs of all possible aggregate snapshots are cached for subsequent dynamic programming recursive calculations.

[0086] S313: Perform dynamic programming recursive calculation.

[0087] First, execute the outer loop: iterate through the atomic snapshot endpoint j of the current interval from 0 to M-1; then, execute the middle loop: iterate through the atomic snapshot endpoint j of the current interval. Constrained by the maximum time slice span, The traversal range is pruned to If the current interval is found at j... In-snapshot cost If the value is infinity (i.e., connectivity breakage has been triggered), then skip the current i to avoid invalid computation. direct assignment ;like The inner loop is then executed: iterating through the cutoff point k of the previous snapshot. Similarly, the range of the previous snapshot is also limited, and the traversal range of k is pruned. arrive Calculate the cost of switching between snapshots within the effective range. The cost of switching between snapshots, also known as the normalized path switching overhead, is calculated using the following formula:

[0088]

[0089] In the formula, It is the shortest path delay from point s to point d in the logical topology of the previous snapshot. It represents the shortest path latency from node s to d in the logical topology of the current snapshot. The normalized path switching overhead means the percentage increase in average service latency across the entire network caused by snapshot merging.

[0090] After calculating the cost of switching between snapshots, perform a state transition and update. Minimum value: And record the optimal predecessor. .

[0091] S314: Construction of the global optimal solution.

[0092] After synthesizing the costs within the snapshot, the final state that minimizes the total cycle cost can be obtained: Record the starting point of the last optimal atomic snapshot. Using the predecessors of the record. Reconstructing the optimal set of interval sequences: First, initialize the set of interval sequences. Then let the starting point of the current interval be... , the previous point Execute the loop until... Find the starting point of the previous interval. ,renew , ,Will Add the interval sequence to the set. Output the set of interval sequences after reversing them. .

[0093] S315: Atomic snapshot merging.

[0094] Traversing the optimal interval sequence set Snapshot the interval of each atom in the set Merge into a single aggregate snapshot to build a logical topology. .

[0095] By implementing the snapshot partitioning method based on state extension dynamic programming described above, the ground station's topology management device can calculate a global snapshot scheduling table that covers the entire system cycle and has the lowest overall cost. This scheduling table contains a series of optimal aggregate snapshot definitions and their corresponding start and end times. However, offline static planning alone is insufficient to cope with the complexity of low-Earth orbit satellite networks in actual operation. On the one hand, the snapshot scheduling table needs to be reloaded on satellite nodes so that the routing execution devices of the satellite nodes can perform route calculations and forwarding based on the scheduling table; on the other hand, sudden link failures beyond ephemeris predictions inevitably occur in the network. Therefore, this application further proposes a distributed routing method based on a full-network logical topology mask, aiming to use the generated global snapshot scheduling table as a benchmark to distinguish between predictable and unpredictable topology changes on satellite nodes, thereby achieving efficient route maintenance.

[0096] S32: Distribute and load the global snapshot schedule to each satellite node.

[0097] In this embodiment, the global snapshot scheduling table is generated by the ground station's topology management device and encapsulated as a configuration update message. This message is then uploaded to the satellite nodes via the power supply link. The satellite nodes flood the configuration update message across the entire network via the laser inter-satellite link. Other satellite nodes, upon receiving the configuration update message, parse out the global snapshot scheduling table and load it into their local memory. For example, Figure 6 This diagram illustrates the generation, distribution, and loading process of the global snapshot scheduling table. Please refer to it. Figure 6 The snapshot partitioning method in step S31 is executed by the topology management device deployed at the ground station. Based on the input satellite constellation ephemeris data, this device calculates a global snapshot scheduling table covering one system cycle and generates a version number for the scheduling table. This scheduling table contains an ordered sequence of aggregated snapshots, and each entry specifically includes: snapshot index k, start time... Termination time And the corresponding logical topology definition. All ground stations encapsulate the generated global snapshot scheduling table into a configuration update message, the header of which carries the version number of the scheduling table. Subsequently, the ground station uploads the configuration message to the satellite nodes within its line of sight via the feed link. After receiving the configuration update message uploaded by the ground station, the satellite node initiates a reliable flooding mechanism, using the laser inter-satellite link to forward the configuration update message to all directly connected neighbor nodes. When any satellite node receives a configuration update message, it first checks the version number in the message header. If the version number is less than or equal to the locally stored version number, it means that the received configuration update message is outdated or duplicated, and it is discarded directly without processing or forwarding; if the version number of the configuration update message is greater than the local version number, it means that it is a new configuration update message, and it needs to be forwarded and processed. To prevent loops during the flooding of configuration update messages, the satellite follows the following principle when forwarding newly received configuration update messages: it does not send the message back to the neighbor node that sent the message, but only forwards it to all other connected neighbor nodes except for the receiving interface. Through the aforementioned reliable flooding mechanism, configuration update messages are propagated along the laser inter-satellite links to the entire network, covering all satellites in the constellation. Upon receiving the configuration update message, the routing execution unit of each satellite node decompresses and verifies it, and stores the parsed global snapshot scheduling table in its local non-volatile memory.

[0098] Furthermore, when the constellation structure undergoes substantial changes (such as the insertion of new satellites into orbit or the decommissioning of old satellites), the ground topology management device will update the local ephemeris database and re-execute the snapshot partitioning method based on the latest constellation configuration, generate a new global snapshot scheduling table and increment its version number, and then repeat the distribution and loading processes.

[0099] S33: Satellite nodes construct and maintain a network-wide logical topology mask based on the logical topology defined in the current snapshot in the global snapshot scheduling table. The logical topology mask records the logical state of all inter-satellite links in the network.

[0100] In this embodiment, satellite nodes construct a network-wide logical topology mask in their local memory based on the logical topology defined in the first snapshot in the snapshot scheduling table. This mask serves as the core data structure for routing decisions. The network-wide logical topology mask is a storage structure that maps all inter-satellite links in the network to their logical states. To enable differentiated handling of predictable and unpredictable changes, the logical topology mask defines three mutually exclusive link state machines: connected state, shielded state, and fault state. Figure 7 A schematic diagram of three link state machine transitions is shown.

[0101] The "Connected" status indicates that the link is active in the current snapshot plan and physical layer communication is normal. In this state, the link is included in the routing calculation topology, and its metrics are effective.

[0102] The "Masked" state indicates that the link is marked as temporarily unavailable in the current snapshot plan. In this state, regardless of whether the physical link is actually connected or disconnected, the routing execution device treats it as disconnected, does not participate in route calculation, and suppresses the link's state updates to achieve zero signaling overhead processing for predictable topology changes.

[0103] The "Failed" state indicates that the link should have been connected in the snapshot planning, but was interrupted due to a sudden failure. In this state, the link does not participate in route calculation, and the generation and elimination of this state correspond to the flooding of LSAs.

[0104] S34: The logical topology mask is maintained through a combination of time-triggered and event-triggered mechanisms.

[0105] In this embodiment, the routing execution unit of the satellite node dynamically maintains the real-time status of the entire network logical topology mask through three processes: time-triggered, local event-triggered, and remote event-triggered.

[0106] For example, Figure 8 A schematic diagram of the time-triggered process is shown. For example... Figure 8 As shown, when the system clock reaches the snapshot switching time... At that time, the routing execution unit reads the topology definition of the next snapshot from the global snapshot scheduling table, queries the status of all links in the logical topology mask, and if the link status is Failed, it means that the unpredictable fault still exists in the new time slice, and the corresponding link status in the logical topology mask remains Failed; if the link status in the logical topology mask is Masked or Connected, the link status of the logical topology mask is set to be consistent with the snapshot topology.

[0107] Figure 9 A diagram illustrating the local event triggering process is shown. For example... Figure 9As shown, for the physical links directly connected to this node, the routing execution unit updates the logical topology mask based on the status changes reported by the physical layer. If the link state sensing device detects a change in the status of a physical link, it queries the status of the corresponding link in the logical topology mask. If the link state of the logical topology mask is Masked, it is determined to be an expected change or a disturbance in the link state. The change is masked, the Masked state remains unchanged, and the generation of link state update messages is suppressed. If the link state of the logical topology mask is Connected or Failed, it is determined to be an unpredictable link disconnection or recovery. The link state of the logical topology mask is updated to be consistent with the physical link state, and a link state update message is generated and flooded throughout the network.

[0108] Figure 10 A schematic diagram of the remote event triggering process is shown. For example... Figure 10 As shown, when a satellite node receives a link state update message from a neighbor's flooded message, it updates its local logical topology mask. If the logical topology mask status of the corresponding link is Masked, the link state update message is masked, and the mask remains unchanged; if the logical topology mask status of the corresponding link is Connected or Failed, the logical topology mask is updated to match the link state update message, and the message is flooded to other neighbors.

[0109] S35: Calculate the routing table for links in the connected state in the logical topology mask and send it to the data plane.

[0110] In this embodiment, satellite nodes trigger route calculation when the overall network logical topology mask undergoes a substantial change. For example, Figure 11 A schematic diagram illustrating the calculations performed by the routing execution device is shown. For example... Figure 11 As shown, the calculation unit calculates the shortest path tree based on the logical topology defined in the current snapshot and the overall network logical topology mask. Only links with a "Connected" status are considered valid edges and participate in path calculation; links with a "Masked" or "Failed" status are considered disconnected during calculation. A routing table is generated based on the shortest path tree and distributed to the forwarding unit in the data plane.

[0111] Through the above steps, this embodiment implements a time-varying routing mechanism of "offline planning guided online execution" in a distributed satellite network. It loads the snapshot scheduling table calculated offline on the ground into the satellite nodes, guiding the routing execution devices of the satellite nodes to make routing decisions. Through this mechanism, satellite nodes can respond to and process predictable topology changes with zero overhead, while simultaneously sensing unpredictable topology changes in a timely manner, updating their local logical topology mask, and flooding it to other nodes to maintain the consistency of the entire network topology.

[0112] The above describes the time-varying routing method for satellite networks provided in this application. By using state-extended dynamic programming for snapshot partitioning, a snapshot scheduling table that balances routing stability and link utilization is obtained. This table is then combined with a logical topology mask to construct a dual-trigger distributed routing mechanism: a time-triggered mechanism enables zero-signaling transition for network snapshot switching; and an event-triggered mechanism responds to and floods unpredictable topology changes. Compared to existing technologies, the low-Earth orbit satellite network snapshot partitioning method based on state-extended dynamic programming avoids excessive snapshot partitioning by balancing the routing update overhead of the control plane and the link utilization of the data plane, thus reducing onboard computational load. Simultaneously, by penalizing severe fluctuations in network latency, the generated snapshot sequence smooths network topology evolution, effectively preventing TCP RTT estimation failures and misadjustments of congestion windows. This, in turn, effectively improves the actual effective throughput of the network while adapting to non-uniform topology changes. The dual-trigger distributed routing method based on the whole network logical topology mask distinguishes between predictable and unpredictable topology changes, achieving zero signaling overhead response to predictable topology changes while retaining the ability to perceive and repair unpredictable topology change events. Thus, a satellite network routing mechanism with both high efficiency and high robustness is realized in a distributed architecture.

[0113] It is understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. Furthermore, in some possible implementations, each step in the above embodiments may be selectively executed according to actual circumstances; it may be partially or fully executed, without limitation here. All or part of any feature of any embodiment of this application can be freely and arbitrarily combined without contradiction. The combined technical solutions are also within the scope of this application.

[0114] Based on the methods in the above embodiments, this application also provides a time-varying routing device for satellite networks. For example, Figure 12 A schematic diagram of a time-varying routing device for a satellite network is shown. Figure 12 As shown, the satellite network time-varying routing device 1200 includes: a snapshot partitioning module 1201, a distribution and loading module 1202, a logical topology mask construction module 1203, a maintenance module 1204, and a route calculation module 1205.

[0115] Among them, the snapshot partitioning module 1201 is used to divide the system cycle into atomic snapshot sequences, and to perform optimal partitioning of the atomic snapshot sequences through the state extension dynamic programming algorithm to generate a global snapshot scheduling table containing aggregate snapshot sequences and their corresponding start and end times;

[0116] The distribution and loading module 1202 is used to distribute and load the global snapshot scheduling table to each satellite node.

[0117] The logical topology mask construction module 1203 is used to construct and maintain a network-wide logical topology mask based on the logical topology defined in the current snapshot in the global snapshot scheduling table. The logical topology mask records the logical status of all inter-satellite links in the network.

[0118] Maintenance module 1204 is used to maintain the logical topology mask in a coordinated manner through time-triggered and event-triggered mechanisms;

[0119] The routing calculation module 1205 is used to calculate a routing table based on the links in the logical topology mask whose link status is connected and send it to the data plane.

[0120] It should be understood that the above-described device is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the device are similar to those described in the above methods. The working process of the device can be referred to the corresponding process in the above methods, and will not be repeated here.

[0121] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0122] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0123] It is understood that the processor in the embodiments of this application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor may be a microprocessor or any conventional processor.

[0124] The method steps in the embodiments of this application can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.

[0125] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0126] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application.

Claims

1. A time-varying routing method for satellite networks, characterized in that, Routing applied to satellite networks, the method includes: The system cycle is divided into atomic snapshot sequences. The atomic snapshot sequences are optimally segmented using a state extension dynamic programming algorithm to generate a global snapshot scheduling table containing aggregate snapshot sequences and their corresponding start and end times. The global snapshot scheduling table is distributed to each satellite node and loaded; Satellite nodes construct and maintain a network-wide logical topology mask based on the logical topology defined in the current snapshot in the global snapshot scheduling table. The logical topology mask records the logical state of all inter-satellite links in the entire network. The logical topology mask is maintained collaboratively through time-triggered and event-triggered mechanisms. Based on the links in the logical topology mask whose link status is connected, a routing table is calculated and sent to the data plane.

2. The method according to claim 1, characterized in that, The process of dividing the system cycle into atomic snapshot sequences, performing optimal segmentation of the atomic snapshot sequences using a state-extended dynamic programming algorithm, and generating a global snapshot scheduling table containing aggregated snapshot sequences and their corresponding start and end times includes: The system period is divided into M equal-length atomic snapshots, forming an atomic snapshot sequence; In the index space of the atomic snapshot sequence In the given information, determine an ordered set of interval coordinate pairs. ; The atomic snapshots within each interval are merged into an aggregate snapshot to generate an aggregate snapshot set. The logical topology corresponding to each element in the aggregate snapshot set is composed of the intersection of the topologies of all atomic snapshots within the interval. Using minimizing the total cost of the entire system as the objective function, the optimal set of interval sequences and the number of segments are solved by the state extension dynamic programming algorithm.

3. The method according to claim 2, characterized in that, The objective function is: In the formula, It is an interval The total cost of merging atomic snapshots into a single aggregate snapshot consists of two parts: routing update cost and performance distortion cost.

4. The method according to claim 3, characterized in that, The cost of route update includes computational overhead. Storage overhead Path switching overhead The performance distortion comes at the cost of wasted link capacity. and delay penalty The formula for calculating the total cost is as follows: In the formula, It is a first-level weighting factor and , It is a second-order weighting factor and satisfies , .

5. The method according to claim 4, characterized in that, The path switching overhead is defined as the cumulative sum of the absolute changes in the shortest path delay of all node pairs in the entire network when switching from the previous aggregate snapshot to the current aggregate snapshot.

6. The method according to claim 2, characterized in that, The state expansion dynamic programming algorithm includes: definition This indicates the atomic snapshot range. The data is divided into segments, and the last aggregated snapshot is the merged interval. Given an atomic snapshot, the global minimum normalization total cost is achieved. Establish the transition state equation: In the formula, is In-snapshot cost, To merge intervals Switch the aggregated snapshot to the merged range The path switching overhead of aggregated snapshots; The final state that minimizes the total cost of the entire cycle is obtained through recursive calculation, and the optimal interval sequence set is reconstructed using the predecessors of the records.

7. The method according to claim 2, characterized in that, The generation of a global snapshot scheduling table containing aggregated snapshot sequences and their corresponding start and end times also includes: Limit the maximum time slice span of aggregated snapshots This constrains the search space of dynamic programming within a sliding window; During the process of traversing and building the aggregate snapshot, the overall network connectivity of the intersection topology is monitored in real time. When a physical disconnection is detected, the subsequent extended traversal under the current starting point is stopped immediately.

8. The method according to claim 1, characterized in that, The time-triggered mechanism includes: When the system clock reaches the snapshot switching time, the satellite node reads the topology definition of the next snapshot from the global snapshot scheduling table; Query the status of all links in the logical topology mask. If the link status is faulty, keep the fault status unchanged; if the link status is shielded or connected, set the link status of the logical topology mask to be consistent with the snapshot topology.

9. The method according to claim 1, characterized in that, The event triggering mechanism includes local event triggering and remote event triggering: Local event triggering refers to querying the status of the corresponding link in the logical topology mask when the status of the physical link directly connected to this node changes. If the link is in a masked state, the change is masked and the generation of link status update messages is suppressed. If the connection is established or the connection is faulty, the logical topology mask is updated and a link state update message is generated and flooded to the entire network. Remote event triggering refers to querying the status of the corresponding link in the logical topology mask when a link status update message flooded by a neighbor is received; if the link is in a blocked state, the message is blocked. If the connection is established or the connection is faulty, the logical topology mask is updated and the message is flooded to other neighbors.

10. A time-varying routing device for satellite networks, characterized in that, The device includes: The snapshot partitioning module is used to divide the system cycle into atomic snapshot sequences. It uses a state extension dynamic programming algorithm to perform optimal partitioning of the atomic snapshot sequences and generate a global snapshot scheduling table containing aggregate snapshot sequences and their corresponding start and end times. The distribution and loading module is used to distribute the global snapshot scheduling table to each satellite node and load it. The logical topology mask construction module is used to construct and maintain a network-wide logical topology mask based on the logical topology defined in the current snapshot in the global snapshot scheduling table. The logical topology mask records the logical state of all inter-satellite links in the network. The maintenance module is used to maintain the logical topology mask in a collaborative manner through time-triggered and event-triggered mechanisms; The routing calculation module is used to calculate a routing table based on the links in the logical topology mask whose link status is connected and send it to the data plane.

Images