Onboard Autonomous Service Orchestration Method for Inter-Satellite Network Splitting and Reconnection

By monitoring ground control heartbeats and inter-satellite adjacency relationships, and utilizing state snapshots and causal version vectors for autonomous decision-making and conflict resolution, the problem of autonomous service orchestration in inter-satellite network splitting and reconnection scenarios has been solved, achieving reliable and secure autonomous service orchestration.

CN123123946APending Publication Date: 2026-11-17HARBIN INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611311945.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-27
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the scenario of inter-satellite network splitting and reconnection, existing technologies lack reliable autonomous service orchestration methods, resulting in high false triggering rates, excessive resource commitments, and difficulty in handling mapping conflicts during reconnection.

Method used

By monitoring ground control heartbeats and inter-satellite adjacency relationships, using state snapshots and causal version vectors for autonomous decision-making and conflict resolution, and employing restricted autonomy and progressive recovery mechanisms, the reliability and security of autonomous behavior are ensured.

Benefits of technology

It achieves reliable autonomy during inter-satellite network splitting and reconnection, reduces the probability of false triggering, avoids excessive resource usage, ensures that the mapping and merging process after reconnection is predictable and controllable, and avoids global service interruption.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123946A_ABST
    Figure CN123123946A_ABST
Patent Text Reader

Abstract

This invention relates to an on-board autonomous service orchestration method for inter-satellite network splitting and reconnection, belonging to the technical fields of satellite network autonomous control, fault-tolerant service orchestration, and distributed state consistency. Specifically, it addresses the technical problems of existing technologies in inter-satellite network splitting and reconnection scenarios, where the orchestration of on-board autonomous services lacks reliable joint splitting determination, restricted autonomy, and deterministic conflict resolution mechanisms, leading to high false triggering rates, over-commitment of resources, and difficulty in handling mapping conflicts during reconnection. The method includes: after detecting reconnection, exchanging snapshot summaries and change logs of each partition, and updating and handling concurrent conflicts according to the causal version vector identification order. This on-board autonomous service orchestration method for inter-satellite network splitting and reconnection is applicable to application scenarios requiring reliable autonomous orchestration in network splitting and reconnection scenarios.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of satellite network autonomous control, fault-tolerant service orchestration, and distributed state consistency, and particularly to on-board autonomous service orchestration methods, apparatus, electronic devices, and computer-readable storage media for inter-satellite network splitting and reconnection. Background Technology

[0002] In a satellite network, satellites are interconnected via inter-satellite links and maintain periodic communication with the ground control center (i.e., "ground control heartbeat"), thus forming a unified orchestration and management system. The ground orchestration controller is responsible for maintaining the Service Function Chain (SFC) mapping, resource allocation, and routing decisions for the entire network, while satellite nodes execute specific Virtual Network Function (VNF) deployments and migrations according to ground instructions.

[0003] However, due to the complexity of the space environment (such as solar storms and electromagnetic interference) or sudden failures of ground facilities, ground control links may be interrupted for extended periods, and inter-satellite links may also be intermittently lost due to satellite movement or equipment malfunctions. In such cases, the originally unified satellite network (constellation) may split into several mutually inaccessible partitions, and global orchestration commands cannot be delivered to the satellite nodes within each partition.

[0004] Existing technologies have the following drawbacks when dealing with the aforementioned network splitting scenarios:

[0005] First, if the partition completely stops orchestration operations, critical services (such as command and control links and emergency communications) will not be able to adapt to changes in local resources, which may lead to a decline in service quality or even service interruption.

[0006] Second, if each partition is allowed to arrange its own configurations without restrictions, problems may arise such as multiple partitions making duplicate commitments to the same satellite resources or making conflicting modifications to the same SFC mapping. This could lead to serious faults such as mapping version conflicts, duplicate VNF instances, and traffic loops during reconnection.

[0007] Therefore, an orchestration method is needed that enables reliable autonomy during network splits and seamless merging after reconnection. This method needs to meet the following requirements:

[0008] It can accurately determine the occurrence of network splits and avoid false triggering caused by instantaneous link jitter;

[0009] It can effectively constrain autonomous behavior during periods of division and prevent over-commitment of resources;

[0010] It can deterministically distinguish between sequential updates and concurrent conflicts after reconnection and restore the unified mapping in a controllable manner. Summary of the Invention

[0011] This invention proposes an on-board autonomous service orchestration method for inter-satellite network splitting and reconnection, which solves the technical problems of existing technologies in inter-satellite network splitting and reconnection scenarios, where on-board autonomous service orchestration lacks reliable joint splitting determination, restricted autonomy and deterministic conflict resolution mechanisms, resulting in high false triggering rate, over-commitment of resources and difficulty in handling mapping conflicts during reconnection.

[0012] The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection described in this invention includes the following steps:

[0013] Monitor the ground control heartbeat, inter-satellite adjacency relationships, and the set of reachable nodes in the control domain;

[0014] When the ground control heartbeat is continuously missing and the change in the set of reachable nodes in the control domain satisfies the splitting condition, the network split is confirmed, and a partition epoch identifier is generated to mark the current autonomous period.

[0015] Verify and load a status snapshot containing the service function chain mapping version, node resource balance, resource lease and risk budget, and determine the partition autonomy quota based on the status snapshot;

[0016] Within the autonomous quota of the partition, restricted virtual network functions are deployed or migrated for critical business operations, and a change log containing the partition epoch identifier, causal version vector, resource lease and rollback point is generated for each mapping change;

[0017] Upon detection of reconnection, the snapshot summaries and change logs of each partition are exchanged, and updates and concurrent conflicts are identified according to the causal version vector.

[0018] For the aforementioned concurrent conflicts, a retention mapping is determined based on lease validity, business priority, and risk budget consumption;

[0019] Establish target instances and paths corresponding to the reserved mapping, switch traffic in batches, and release conflicting resources.

[0020] Furthermore, a preferred embodiment is provided, wherein the splitting condition includes: the number of consecutive missing ground control heartbeats reaches a first threshold, and the symmetric difference between the current control domain reachable node set and the previous control domain reachable node set reaches a second threshold.

[0021] Furthermore, a preferred embodiment is provided in which the partition epoch identifier includes a split event sequence number, a partition representative node identifier, and a time base, used to distinguish changes generated during different autonomous periods.

[0022] Furthermore, a preferred implementation is provided, wherein the state snapshot further includes authorized quota, critical business list, resource reservation amount, nearest feasible path and integrity verification value; when the integrity verification value indicates that the integrity verification of the state snapshot fails, the partition enters minimal autonomy mode, wherein the minimal autonomy mode maintains the existing instance and prohibits the addition of new low-priority instances.

[0023] Furthermore, a preferred embodiment is provided, wherein the process of determining the autonomous quota for the partition includes:

[0024] The recommended amount is calculated based on the valid resource leases, key business priorities, and risk budgets in the aforementioned status snapshot;

[0025] The suggested quota is compared with the authorized quota in the status snapshot and the value of the available idle resources in this partition after deducting the fault reservation amount, and the minimum value is taken as the final partition autonomy quota.

[0026] Furthermore, a preferred implementation is provided in which the restricted virtual network function deployment or migration meets the following restrictions: operations are performed only on service function chains with a priority not lower than a preset level; the number of migrations for a single service function chain does not exceed a preset number; the total amount of resources consumed by all operations does not exceed the partition autonomy quota; and only resources authorized by valid resource leases that have not yet expired are used.

[0027] Furthermore, a preferred implementation is provided, wherein the batch switching of traffic includes: the proportion of each batch of traffic does not exceed a preset proportion, and the health status, end-to-end latency and queue growth rate of the target instance are verified before the next batch switching; if the verification fails, the subsequent switching is stopped and the original mapping is restored according to the rollback point.

[0028] This invention also proposes an on-board autonomous service orchestration method and apparatus for inter-satellite network splitting and reconnection, the apparatus comprising the following modules:

[0029] The status monitoring module is used to monitor the ground control heartbeat, inter-satellite adjacency relationships, and the set of reachable nodes in the control domain;

[0030] The split confirmation module is used to confirm network splitting when the ground control heartbeat is continuously missing and the change in the set of reachable nodes in the control domain satisfies the splitting condition, and to generate a partition epoch identifier.

[0031] The snapshot verification module is used to verify and load a status snapshot that includes the service function chain mapping version, node resource balance, resource lease and risk budget.

[0032] The restricted autonomy module is used to determine the partition autonomy quota based on the state snapshot, and to perform virtual network function deployment or migration within the partition autonomy quota;

[0033] The causal log module is used to generate change logs for mapping changes during splits, including the partition epoch identifier, causal version vector, resource lease, and rollback point.

[0034] The conflict resolution module is used to exchange snapshot summaries and change logs of each partition after reconnection, identify sequential updates and concurrent conflicts based on the causal version vector, and determine the retention mapping according to lease validity, business priority and risk budget consumption.

[0035] The progressive recovery module is used to establish target instances and paths corresponding to the retained mapping, switch traffic in batches, perform health verification and release conflicting resources, and restore the original mapping according to the rollback point when the verification fails.

[0036] Furthermore, in a preferred embodiment, the snapshot verification module is also used to trigger a minimum autonomy mode when the integrity verification of the state snapshot fails. The minimum autonomy mode maintains the existing instance and prohibits the addition of new low-priority instances. The conflict resolution module is also used to select the reserved mapping according to the deterministic node identifier order when the lease validity, business priority and risk budget consumption are all the same.

[0037] The present invention also proposes an electronic device comprising: a processor and a memory, the memory being used to store executable instructions of the processor, the processor being configured to execute, via executing the executable instructions, the on-board autonomous service orchestration method for inter-satellite network splitting and reconnection as described above.

[0038] The present invention has the following beneficial effects:

[0039] 1. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection described in this invention, by using a joint determination of the number of consecutive missing ground control heartbeats and the symmetry difference of the set of reachable nodes in the control domain, and setting a confirmation window for secondary confirmation, can effectively distinguish between real network splitting and instantaneous link jitter, significantly reducing the probability of false triggering caused by a single short-term packet loss or topology fluctuation, thereby improving the accuracy of autonomous operation initiation.

[0040] 2. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection described in this invention introduces resource leases, critical service priorities, and risk budgets into state snapshots, calculates the autonomous quota for each partition based on these three factors, and then uses the snapshot authorization quota and idle resources minus fault retention as hard capping constraints. This can precisely limit the resource consumption of VNF deployment or migration during splitting, avoid multiple partitions from repeatedly committing to and overusing the same resource, thereby ensuring the controllability and security of autonomous behavior.

[0041] 3. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection described in this invention uses causal version vectors to record the number of updates to the same mapping object by each partition, and compares the version vectors of each partition during reconnection. This can deterministically distinguish between sequential updates and concurrent conflicts, and then resolve conflicts according to progressive rules based on lease validity, service priority, risk budget consumption, and deterministic node identification order. This makes the mapping merging process after reconnection predictable and reproducible, avoiding the randomness and uncertainty of manual intervention.

[0042] 4. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection described in this invention, by adopting a gradual recovery approach of switching traffic in batches and verifying the health status of target instances, end-to-end latency, and queue growth rate before each batch switch, can limit the impact of potential faults to a single batch of traffic during the reconnection process, avoiding global service interruption caused by full switch failure. At the same time, combined with the rollback point mechanism, it can quickly restore to the original mapping when verification fails, thereby achieving a smooth transition and high availability in the reconnection process.

[0043] The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection described in this invention is applicable to application scenarios such as dynamic satellite networks, space-ground integrated information networks, and distributed constellation systems that require reliable autonomous orchestration in network splitting and reconnection scenarios. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments 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 from these drawings without creative effort.

[0045] Figure 1 A flowchart of an on-board autonomous service orchestration method for inter-satellite network splitting and reconnection, as described in one embodiment of the present invention;

[0046] Figure 2 This is a flowchart of network split joint determination in one embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram illustrating the contents of a state snapshot in one embodiment of the present invention.

[0048] Figure 4 This is a flowchart illustrating the orchestration process under inter-satellite network splitting and reconnection in one embodiment of the present invention.

[0049] Figure 5A flowchart illustrating conflict resolution based on causal version vectors is provided in one embodiment of the present invention.

[0050] Figure 6 A flowchart of progressive recovery and unified mapping is provided in one embodiment of the present invention;

[0051] Figure 7 This is a schematic diagram of the functional modules of an on-board autonomous service orchestration device for inter-satellite network splitting and reconnection, as described in one embodiment of the present invention. Detailed Implementation

[0052] To make the technical solutions and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail and completely below with reference to the accompanying drawings. The various embodiments described below are only some preferred embodiments of the present invention, and not all of them; the various embodiments described below are intended to explain the present invention and should not be construed as limiting the present invention; reasonable combinations of the technical features defined in the various embodiments of the present invention, as well as all other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort, are all within the scope of protection of the present invention.

[0053] It should be noted that simply stating optimization goals, adopting a certain general algorithm, or claiming to improve the success rate does not constitute a directly implementable technical solution. Therefore, this invention incorporates the formation method of state data, access and security constraints, abnormal branches, and deployment execution order into the technical process, forming a complete and implementable solution.

[0054] Implementation Method 1: On-board Autonomous Service Orchestration Method for Inter-Satellite Network Splitting and Reconnection:

[0055] Technical Background and Problem Overview: In satellite networks, satellites are connected via inter-satellite links and maintain communication with the ground control center (i.e., "ground control heartbeat"). When the ground control center is unable to contact us for an extended period due to unforeseen circumstances (such as natural disasters or equipment failure), the satellite network may split into several independent sub-networks (called "partitions"). At this time, each partition needs to possess the capability to autonomously manage services (i.e., "on-board autonomy") to ensure uninterrupted critical services. Once communication is restored, these partitions need to be re-merged (i.e., "reconnection"), and service configuration conflicts arising during this period need to be resolved. This method aims to address these issues.

[0056] The method includes the following steps:

[0057] Step 1: Continuous monitoring and split prediction:

[0058] Step 1.1: Continuously monitor the following status data:

[0059] Ground control heartbeat: A liveness signal periodically sent by the ground control center to the satellite to prove that the ground control center is still functioning normally; the absence of a heartbeat means that ground control may have failed.

[0060] Inter-satellite adjacency: The connection status (whether communication is possible) between each satellite and its neighboring satellites;

[0061] Set of reachable nodes in the control domain: At the current moment, the set of all other satellites that this satellite can successfully communicate with via inter-satellite links; this set changes as inter-satellite links are disconnected or satellites move.

[0062] Step 1.2: Prediction and Confirmation of Split:

[0063] The system continuously monitors the above three data points. When the following two conditions (split conditions) are met, it is considered that a network split may have occurred:

[0064] ① If the number of consecutive missing heartbeats in ground control reaches a preset threshold (first threshold, for example, 3 times), it indicates that ground control may have failed.

[0065] ② The current set of reachable nodes in the control domain has changed significantly compared to the set of reachable nodes in the control domain at the previous moment (measured by the "symmetry difference", i.e., the number of different elements in the two sets; if the symmetry difference reaches the second threshold), indicating that the network topology has changed drastically.

[0066] Suspected split state: Once the above split conditions are met, the system immediately enters the "suspected split state", but does not immediately take autonomous action (i.e., start on-board autonomy) to avoid misjudgment due to momentary interference;

[0067] Confirmation Window: This is a preset time period (e.g., 2 heartbeat cycles); within this window, the system continuously monitors whether the above conditions are still met.

[0068] Formal confirmation of network split: If the above split conditions are still met at the end of the confirmation window, the system formally confirms that the network has split.

[0069] Partition Era Identifier: Once the split is officially confirmed, the system immediately generates a globally unique tag to mark all operations during this autonomous period (i.e., the on-board autonomous period); this tag contains the following three parts to distinguish changes generated in different autonomous periods:

[0070] Split event number: an incrementing number used to distinguish different split events in history;

[0071] Partition Representative Node Identifier: The identity information of a specific node within the current partition (e.g., the satellite with the smallest ID), used to distinguish different partitions;

[0072] Time reference: Identifies the moment of splitting and is used to align the timeline.

[0073] Technical Function: By employing dual checks (heartbeat + topology change) and a confirmation window, the probability of false triggers is significantly reduced, ensuring the reliability of autonomous operations. The partition epoch marker provides a timeline and a basis for partition ownership in subsequent conflict resolution.

[0074] In this embodiment, network splitting is determined jointly by ground-controlled heartbeat loss and changes in the set of reachable nodes, avoiding false triggering caused by a single short-term packet loss.

[0075] Step 2: Status Snapshot Verification and Autonomous Quota Determination:

[0076] Step 2.1: State Snapshot:

[0077] During normal operation, the system periodically saves a complete record of the network and service status, which is called a "state snapshot." It's like a "health check report," recording the health status before a split occurs.

[0078] A state snapshot should include at least the following:

[0079] Service Function Chain (SFC) Mapping Version: An SFC is a service chain consisting of multiple Virtual Network Functions (VNFs, such as firewalls and load balancers); the mapping version records which satellite each VNF on this service chain is deployed on; the version number is used to track changes;

[0080] Node resource reserves: The remaining available resources such as CPU, memory, and storage for each satellite;

[0081] Resource lease: A temporary resource authorization contract that specifies the rights, capacity, expiration time, and source of authorization for a VNF to use a specific resource; for example, "VNF-A has the right to use 2 CPU cores on satellite X until time T"; each resource lease contains the following four subfields:

[0082] ①Resource object: The specific resource entity identifier that is authorized, such as "CPU cores 0-3 of satellite X" or "100GB of space in storage volume Y";

[0083] ② Authorized capacity: The amount of resources allowed to be used under this lease, such as "2 CPU cores" or "50GB of memory";

[0084] ③ Expiration Slot: The valid expiration time stamp or slot number of the lease; the lease will automatically expire after this time.

[0085] ④ Authorization Era: The version / round number of the lease, used to distinguish different leases for the same resource in different periods to avoid conflicts;

[0086] ⑤ Lease expiration release rule: When the system clock reaches the lease expiration time slot and the lease has not been renewed, the corresponding temporary resource is automatically marked as "reclaimable" and returned to the idle resource pool in the next resource cleanup cycle; this mechanism ensures that temporary resources are not ineffectively occupied for a long time;

[0087] Critical business list: Identifies which SFCs are of the highest priority (such as command and control links) and must be prioritized for protection;

[0088] Risk budget: The maximum risk the system is allowed to assume during autonomy (e.g., how many low-priority services can be interrupted at any given time).

[0089] Integrity check value: A checksum (such as a hash value) used to verify that the state snapshot file has not been corrupted during transmission or storage.

[0090] The state snapshot also includes authorized limits, resource reservations, and the nearest feasible path:

[0091] Authorization limit: The upper-level controller or cluster master controller pre-grants the upper limit of resources (an absolute value) for this partition, which is used to limit the total amount of resources consumed during the autonomous period;

[0092] Resource reservation: The amount of resources that have been reserved but not yet allocated (e.g., spare capacity reserved for critical business operations).

[0093] Nearest Feasible Path: The set of optimal route alternatives for each SFC under the current network topology, used to quickly restore communication.

[0094] Step 2.2: Verification and Loading (in memory):

[0095] After the split, the system first finds and loads the most recently saved state snapshot and verifies its integrity.

[0096] Verification failed: If the snapshot is corrupted, the system cannot know the exact state before the split, so it enters the "minimal autonomy mode": only the existing running VNF instances are maintained (referred to as existing instances, which are not created, migrated or destroyed), and the creation of new VNFs or the migration of any existing VNFs are strictly prohibited to prevent destructive decisions due to incorrect state information and to ensure that existing services do not crash.

[0097] Verification passed: The system extracts all valid resource leases, critical business priorities, and risk budgets from the status snapshot for calculating the autonomous quota for the partition.

[0098] ①Valid resource leases: This refers to the list of resource leases that have not yet expired. The sum of the authorized capacities of these leases constitutes the resource base that can be legally used in this partition;

[0099] ② Prioritize key business operations: Extract the priority level of each SFC from the list of key business operations; the minimum resource amount required for high-priority business operations should be reserved in the quota calculation.

[0100] ③ Risk Budget: The maximum risk limit that the system allows to be consumed during this autonomous period, used to adjust the allowed excess commitment ratio in the limit calculation.

[0101] In addition, the system also extracts the authorized quota from the state snapshot, which is used as one of the hard limits in subsequent calculations.

[0102] Step 2.3: Regional Autonomy Quota:

[0103] The autonomous zone limit is the maximum amount of resources that a zone can utilize during its autonomous period. Its determination process involves two steps:

[0104] Step 1: Calculate the "recommended limit" based on business strategy:

[0105] Based on the valid resource leases, key business priorities, and risk budget extracted in step 2.2, the system calculates a theoretically recommended limit; the calculation method is as follows:

[0106] Valid resource leases: The sum of the authorized capacity of all leases that have not yet expired in the current partition, denoted as Base (resources corresponding to expired leases are not included);

[0107] Critical business priority: Resources required for high-priority businesses must be guaranteed first; if the minimum amount of resources required for high-priority businesses is Reserved_High, then the basic amount of resources available for other operations is Base_avail = Base - Reserved_High (if Base < Reserved_High, then Base_avail = 0).

[0108] Risk budget: Let the risk budget be R (for example, 0.2 means that an over-limit of 20% is allowed), then the suggested limit is Suggested = Base_avail × (1 + R).

[0109] Step 2: Apply a hard cap to determine the final credit limit:

[0110] The partition autonomy quota shall not exceed the smaller of the following two values: "the authorized quota recorded in the status snapshot" and "the currently confirmed idle resources of this partition minus the resources that must be reserved for fault recovery"; wherein:

[0111] The currently verifiable available free resources in this partition are: the total remaining resources of all satellites within the partition as obtained through real-time detection (excluding those already leased).

[0112] Fault reserve: The amount of resources (e.g., 10% of total resources) that are forcibly reserved to cope with possible future satellite failures or link interruptions, and shall not be used for regular scheduling;

[0113] Specifically, "State snapshot authorization quota (denoted as Q_snap)" and "Identifiable free resources - fault reservation amount for this partition (denoted as Q_free-Reserve)" are two hard upper limits of partition autonomy quota.

[0114] Compare the "suggested limit" obtained in the first step with the two hard limits mentioned above, and take the minimum value as the final autonomous limit for the region:

[0115] The above formula ensures that the autonomous quota does not exceed the authorization of the superior and does not overdraw future fault redundancy. This can prevent the overuse of resources and leave room for possible future failures.

[0116] Step 3: Restricted autonomy within a partition:

[0117] Step 3.1: Organize within the partition:

[0118] After the split is confirmed, the satellites in the partition need to select a temporary "person in charge" (temporary coordinating node), or if there is no person in charge, everyone follows a set of preset local rules (masterless local rules) to make decisions together.

[0119] Step 3.2: Perform new deployments or migrations of Virtual Network Functions (VNFs) with restricted critical business operations:

[0120] The core of autonomy is to make necessary adjustments to key operations when resources are limited; all operations must simultaneously meet the following four constraints:

[0121] ①Priority restriction: Operations can only be performed on SFCs with a priority level no lower than the preset level; low-priority services are temporarily frozen.

[0122] ② Migration Limit: A single SFC has a limit on the number of migrations during the entire autonomous period (e.g., it can only be migrated once at most) to avoid instability caused by frequent changes;

[0123] ③ Available capacity limit: The total amount of resources consumed by all operations cannot exceed the “partition autonomy quota” (i.e., the available capacity limit) determined in step 2.

[0124] ④ Lease expiration restrictions: Resources can only be used under valid leases that have not yet expired (subject to the lease expiration date).

[0125] Technical principle: Through strict restrictions, the system's behavior can still be controlled and stable even when ground control is lost, avoiding situations of resource exhaustion or service avalanche.

[0126] Step 4: Change log generation:

[0127] Change Log: During a split, every modification to the SFC mapping (such as migrating a VNF from satellite A to satellite B) generates a detailed log entry. This is like an "operation logbook".

[0128] The log should include at least the following:

[0129] Partition Era Identifier: Indicates which split event this operation occurred in;

[0130] Object Identifier: A unique ID of the modified SFC or VNF;

[0131] Causal version vector: Each component records the number of times each partition updates the same mapped object; specifically, it is an array, where each dimension of the array represents a partition, and the value indicates how many times the object has been modified in that partition; for example, [1, 0, 2] means that partition 1 was modified once, partition 2 was not modified, and partition 3 was modified twice; this vector is the key to resolving conflicts.

[0132] Resource leases: Lease information used in this operation;

[0133] Rollback point: Complete state information before the operation, used to restore the state before the modification in case of an error.

[0134] Step 5: Reconnection Detection and Collision Identification:

[0135] Step 5.1: Reconnection Trigger: When ground control is restored or inter-satellite links are reconnected, and the previously split partitions are able to communicate with each other again, the system detects a reconnection event.

[0136] Step 5.2: Suspend Low-Priority Changes: Upon detecting reconnection, the system immediately suspends all new low-priority change operations (such as VNF creation, migration, scaling up, and scaling down for non-critical business operations) to prevent further inconsistencies during conflict resolution. High-priority critical business changes can still be executed according to the risk budget, but they must be specially marked for subsequent processing.

[0137] Step 5.3: Exchange Summary and Log: After reconnection, each partition has its own modified SFC mapping; they need to exchange their respective "State Snapshot Summary" (a simplified version of the snapshot, used for quick comparison) and "Change Log":

[0138] State Snapshot Summary: A simplified version of a snapshot (such as a hash tree or checksum of a key field) used for quick comparison of differences;

[0139] Changelog: A record of all modifications made to objects such as SFC mappings and resource leases in this partition since the last public snapshot, along with a causal version vector for each modification.

[0140] Step 5.4: Causal Version Vector Conflict Identification: The system uses the causal version vector to determine the modification order of each partition. The version vector of each partition is an integer array, and each component of the array records the number of times that partition has updated a certain mapped object (for example, the vector of partition A is [1,0], which means that A has been updated 1 time and B has been updated 0 times).

[0141] Compare the version vectors of any two partitions:

[0142] Sequential update: If a vector can "dominate" another vector (i.e., each component is greater than or equal to the other, and at least one component is strictly greater than it), then the modification of the latter is based on the former, and the latest result can be directly adopted (i.e., the modification of the dominated vector already includes the result of the dominant vector).

[0143] Concurrent conflict: If two vectors do not dominate each other (or the two versions are not comparable, for example, A=[1,0], B=[0,1]: the first component of A is greater than B, but the second component is less than B), it means that both parties have modified the same object in the same time period and are unaware of each other's modifications, thus indicating a concurrent conflict. The next step of the conflict resolution process is required, and the modifications of both parties are merged according to a preset strategy (such as based on priority, timestamp, or manual adjudication).

[0144] Step 6: Conflict resolution and retention mapping determination:

[0145] Step 6.1: Conflict Resolution: For identified concurrent conflicts, it must be decided which partition's modification results to retain; resolution rules are judged in order of priority:

[0146] ①Lease Validity: Which partition's modification is based on a resource lease that is still valid? The valid lease wins.

[0147] ② Business Priority: If all leases are valid, the priority of the business involved in the modification is compared, and the one with higher priority wins;

[0148] ③ Risk budget consumption: If the priorities are the same, compare which partition consumes less risk budget and the partition with less consumption wins;

[0149] ④ Deterministic node identification order (as the final decision condition): If all three items above are the same (i.e., lease validity, business priority, and risk budget consumption are all the same), then the order is determined by comparing the IDs of the representative nodes of the two partitions, with the smaller ID winning.

[0150] Step 6.2: Retain Mapping: The final version selected by the above rules is the "Retain Mapping", which is the correct configuration that should be used after reconnection.

[0151] Step 7: Progressive Recovery and Unified Mapping:

[0152] Step 7.1: Establish target instance and path: Based on the "reserved mapping", prepare the required VNF instance (hereinafter referred to as: target instance) on the corresponding satellite and plan the communication path for data flow.

[0153] Step 7.2: Gradual Switching: In order not to affect the running services, instead of cutting off all the old traffic at once, a "canary release" approach (or a gradual approach to switch traffic in batches) is adopted to switch traffic from the old mapping to the new mapping in batches; for example, the first batch switches 20% of the traffic.

[0154] Step 7.3: Health Verification: Before each batch of traffic switching, a rigorous check must be performed on the newly established instances and paths:

[0155] Target instance health status: Is the new VNF running normally? Is the CPU / memory usage normal?

[0156] End-to-end latency: Whether the delay of data packets traveling along the new path meets the requirements;

[0157] Queue growth rate: Is the data queue length on the new path increasing (if it is increasing too fast, it indicates path congestion);

[0158] After successful verification, continue switching to the next batch until all batches have been switched.

[0159] Step 7.4: Success and Failure Handling:

[0160] Success: If all batches pass verification, all traffic is switched to the new mapping, the conflicting resources occupied by the old mapping are released, and the system officially exits this "partition era" and returns to the normal mode of unified management.

[0161] Failure: If a batch of verifications fails (e.g., a new instance crashes), the subsequent switchover will be stopped immediately, and all traffic will be switched back to the old mapping based on the previously recorded "rollback point" to ensure uninterrupted service.

[0162] In summary, this method achieves reliable, controllable, and lossless on-board autonomous service orchestration in inter-satellite network splitting and reconnection scenarios through a complete closed loop of "monitoring-prediction-snapshot-restricted autonomy-logging-conflict resolution-gradual recovery". Each step includes clear judgment criteria and fault tolerance mechanisms, ensuring the robustness of the entire system.

[0163] Implementation Method 2: Some formulaic relationships in the on-board autonomous service orchestration method for inter-satellite network splitting and reconnection:

[0164] (1) The conditions for confirming the split are expressed as follows:

[0165] h_loss≥H and |R(t)△R(t-1)|≥K;

[0166] Where h_loss is the number of consecutive heartbeats missing, R(t) is the set of reachable nodes, Δ is the symmetric difference, and H and K are the first threshold and the second threshold, respectively.

[0167] (2) The partition autonomy quota is no greater than the "state snapshot authorization quota (denoted as Q_snap)" and the "identifiable free resources - fault reserve for this partition (denoted as Q_free-Reserve)". This is expressed as:

[0168] Q_p min(Q_snap,Q_free-Reserve);

[0169] Where Q_snap is the snapshot authorization quota (the authorization quota recorded in the status snapshot), Q_free is the available free resources in this partition, and Reserve is the fault reservation amount.

[0170] (3) Causal version vector identification sequence update and concurrent conflicts:

[0171] Let the (causal) version vector of object x be v_x;

[0172] If all components of v_a are not greater than v_b and at least one is less than v_b, then a is causally prior to b, which is a sequential update.

[0173] If v_a and v_b do not control each other, then it is a concurrent conflict and enters the conflict resolution stage.

[0174] (4) Gradual batch switching of traffic:

[0175] The proportion of traffic in each batch during the gradual switchover should not exceed γ, and the health of the target instance, end-to-end latency, and queue growth rate should be verified before the next batch switchover.

[0176] In this embodiment, the specific weights, thresholds, and window lengths in the above formula relationships are obtained from configuration files, historical quantiles, or online estimation. Any equivalent implementation can be used as long as the replaced quantities maintain the same physical meaning, monotonic relationship, and constraint effect.

[0177] Implementation Method 3: Data Acquisition and Validation Processing

[0178] This implementation method is a further refinement of steps 1 (continuous monitoring) and 2 (status snapshot verification) in implementation method 1. It mainly explains the source, format, and preprocessing procedures for monitoring data and status snapshots before they are used.

[0179] I. Sources and formats of status data:

[0180] 1. In step 1 of implementation method 1, the system needs to continuously monitor the "ground control heartbeat," "inter-satellite adjacency," and "control domain reachable node set." These data are collectively referred to as state data, and they originate from the following entities:

[0181] Ground orchestration controller: A central control system located on the ground, responsible for sending heartbeat signals and global scheduling commands;

[0182] On-board orchestration agent: A software module running on each satellite, responsible for collecting information on the satellite's neighbors and resource status;

[0183] Network telemetry unit: A specialized hardware or software component used to collect network metrics such as link quality, latency, and bandwidth utilization in real time.

[0184] 2. For status data, each status record (i.e., data obtained from a single sampling) must contain at least the following fields:

[0185] Node identifier: A unique ID (e.g., satellite number) of the satellite or device that generated the data;

[0186] Sampling time slot: The time (time stamp) when the data was collected, used to determine the freshness of the data;

[0187] Data version: The version number of this data, used to track whether the data has been updated;

[0188] Validity period: The length of time (e.g., 30 seconds) during which the data is considered "valid". After this time, the data will be considered expired.

[0189] II. Preprocessing of status data before use:

[0190] Before using state data for decision-making (such as determining whether to split or calculating autonomy limits), the system needs to perform the following preprocessing:

[0191] Timestamp verification: Checks whether the data sampling time slot is within the validity period; if the data is too old (exceeds the validity period), it cannot be used directly;

[0192] Unit normalization: unify the units of data from different sources (e.g., unify memory units to MB, and time units to milliseconds) to avoid calculation errors caused by inconsistent units;

[0193] Outlier handling: Remove obviously erroneous data (such as negative resource reserves or bandwidth values ​​exceeding physical limits) and replace them with interpolated or historical average values.

[0194] III. Handling of Expired Data:

[0195] If a status record exceeds its validity period, it is called expired data. The system will not directly use expired data for decisions in the current time slot (i.e., decisions in this time slot, such as determining whether the splitting condition is met), but will trigger one of the following three remedial measures:

[0196] Remeasure: Immediately initiate a new measurement request to the data source to obtain the latest data;

[0197] Predictive compensation: Using historical data trends or machine learning models, estimate an approximate value for the current moment (e.g., predict whether the current heartbeat is normal based on the heartbeat pattern of the past 5 minutes).

[0198] Safety rollback: If new data cannot be obtained or the situation cannot be predicted, a conservative strategy is adopted (e.g., assuming that the heartbeat is lost or the link is broken) to ensure safety.

[0199] IV. Service Request Transformation:

[0200] In step 3 (Restricted Autonomy) of Implementation 1, the system needs to perform VNF deployment or migration for critical services. These services initially enter the system in the form of service requests. A service request is an instruction initiated by a user or upper-layer application that requests the establishment or modification of a Service Function Chain (SFC).

[0201] When a service request enters the orchestrator (the software module responsible for decision-making), it is converted into a machine-processable data object; this object contains the following information:

[0202] The order of functions or tasks: that is, the execution order of each VNF in the SFC; for example, data must first pass through the firewall, then through the load balancer, and finally reach the application server.

[0203] Resource requirements for each function: How much CPU, memory, storage, and other resources does each VNF require;

[0204] Traffic requirements between adjacent functions: The data bandwidth required to be transmitted between two adjacent VNFs (e.g., 100Mbps from a firewall to a load balancer).

[0205] End-to-end quality of service constraints: The total latency limit, packet loss rate requirements, etc. that the entire SFC must meet.

[0206] V. Topological Structure of Service Function Chain:

[0207] A Service Function Chain (SFC) can be a simple linear chain (A→B→C) or it can be extended to a more complex Directed Acyclic Graph (DAG), which includes a branching and merging structure. For example, after passing through a firewall, data can simultaneously flow to a load balancer and a caching server before converging to the application server. This flexibility allows this method to adapt to various business scenarios.

[0208] Implementation Method 4: Transaction Execution and Exception Branches

[0209] This implementation is a further refinement of steps 3 (restricted autonomy within the partition) and 7 (progressive recovery and unified mapping) in implementation 1. It mainly explains how to ensure the security and atomicity of operations when performing VNF deployment or migration, and how to roll back in abnormal situations.

[0210] I. Basic Concepts of Transaction Execution:

[0211] In step 3 of implementation 1, the system will perform new deployments or migrations of VNFs within the partition; these operations may change the existing SFC mapping, so they must be performed like database transactions, ensuring that they either all succeed or all fail (leaving no unfinished products), which is the meaning of transaction execution.

[0212] II. Generation and review of candidate solutions;

[0213] Any decision to change the existing SFC mapping (such as migrating a VNF from satellite A to satellite B) must go through the following process before it is actually implemented:

[0214] Candidate solution generation: Based on the current resource situation and business needs, the system generates one or more feasible migration or deployment solutions (i.e., candidate solutions); each solution details which VNF should be placed on which satellite, how many resources should be used, and which path should be taken, etc.

[0215] Temporary resource reservation: Reserve the necessary resources in advance for the operations in the candidate scheme (e.g., reserve CPU and memory on the target satellite) to prevent other operations from preempting them;

[0216] End-to-end review: A comprehensive review of candidate solutions to ensure their feasibility; the review includes:

[0217] ① Node capabilities: Does the target satellite have sufficient computing power and storage space?

[0218] ② Computation and storage capacity: Whether the specific values ​​meet the requirements of VNF;

[0219] ③ Link bandwidth: Whether the path from source to destination has sufficient bandwidth;

[0220] ④ Path connectivity: Does a usable communication path exist?

[0221] ⑤ Latency Limit: Whether the latency of the new path is lower than the maximum value required by SFC;

[0222] ⑥ Business continuity: Will the handover process cause service interruption (e.g., zero-interruption handover is required).

[0223] III. Submission and Failure Handling:

[0224] All checks passed: If all checks pass, the plan is officially submitted, and the actual VNF deployment or migration begins.

[0225] If any check fails (e.g., insufficient bandwidth), the previously temporarily reserved resources are immediately released to prevent the partially completed mapping from entering the production state (i.e., partially completed VNFs are not allowed to go online to avoid inconsistencies).

[0226] IV. Rollback mechanism in abnormal situations:

[0227] 1. During execution, the following abnormal situations may be encountered:

[0228] The solution was not completed within the decision deadline: The system failed to calculate the final solution within the specified time (e.g., the algorithm timed out).

[0229] Participating node disconnection: A satellite that needs to participate in the operation suddenly goes offline;

[0230] The state version has changed: During the calculation process, other operations have altered the resource state (e.g., another satellite has occupied the reserved resources).

[0231] 2. When any of the above situations occur, the system will stop submitting the current candidate solution and trigger a rollback mechanism; the rollback methods are as follows:

[0232] Maintain original mapping: Do not make any changes and maintain the SFC mapping before the split.

[0233] Use the most recent feasible mapping: if there was a successful mapping before, revert to that version.

[0234] Pre-registered security actions are only performed on high-priority services: For urgent critical services, pre-set contingency plans can be executed (such as forced switch to backup satellites).

[0235] V. Technical Effects:

[0236] The aforementioned rollback method ensures that even if the optimization algorithm malfunctions (such as an infinite loop or data inconsistency), the normal operation of existing services will not be disrupted. In other words, any failed decision will not cause the running service to be interrupted, thus achieving high availability.

[0237] Implementation Method 5: Node, Link, and Virtual Network Functions in the On-board Autonomous Service Orchestration Method for Inter-Satellite Network Splitting and Reconnection:

[0238] Satellite nodes can be LEO, MEO, or GEO satellites, and can form a candidate resource pool together with ground edge nodes and high-altitude platform nodes.

[0239] The link can be a radio frequency link, a laser inter-satellite link, or a combination of both.

[0240] Virtual networking capabilities can be implemented through containers, virtual machines, programmable data plane capabilities, microservices, or onboard native missions.

[0241] Those skilled in the art can substitute the distance function, predictor, distributed solver, message format, and deployment carrier without altering the data dependencies and security constraints between the core steps; these substitutions still fall under the equivalent implementation of the present invention.

[0242] Implementation Method 6: A specific embodiment of an on-board autonomous service orchestration method for inter-satellite network splitting and reconnection:

[0243] In this embodiment, the ground control heartbeat cycle is one time slot. When there are three consecutive missing heartbeats and the reachable set changes by no less than two nodes, the network split confirmation process is initiated.

[0244] A state snapshot is generated every 10 time slots, and a lease lasts for a maximum of 30 time slots.

[0245] During the split, only SFCs with a priority no lower than the preset level are allowed to be created or migrated; a single SFC can be migrated at most once, and the original mapping rollback point is retained for each change.

[0246] If snapshot verification fails, the system will enter a minimal autonomous mode that only allows existing instances to be maintained and does not allow new instances to be added.

[0247] When reconnecting to the network, critical business conflicts are handled first, followed by ordinary business; a maximum of 20% of traffic is switched in each batch.

[0248] Implementation Method 7: System, Equipment, and Media Implementation Method:

[0249] This embodiment describes the system architecture, hardware device, storage medium, and computer program product for implementing the aforementioned orchestration method (embodiments 1 to 6).

[0250] I. System Execution Environment:

[0251] 1. This method (an on-board autonomous service orchestration method for inter-satellite network splitting and reconnection) can be run in any of the following environments:

[0252] Ground-based data centers: traditional centralized server clusters;

[0253] Constellation Controller: A ground station specifically designed to manage the entire satellite constellation;

[0254] Onboard computing platform for a single satellite: the computer built into the satellite;

[0255] Multi-satellite collaborative platform: A distributed computing environment formed by multiple satellites collaborating through a network.

[0256] 2. The various functional modules of the system can be implemented by independent hardware (such as dedicated chips) or by software components executed by a general-purpose processor; the modules exchange status and decision information in the following ways:

[0257] Message bus: An internal communication channel between modules;

[0258] Inter-satellite link: a wireless communication link between satellites;

[0259] Control interface: The interaction interface between the module and the external controller.

[0260] II. Electronic Equipment:

[0261] An electronic device for performing this method, the basic hardware components of which include:

[0262] At least one processor: This is the brain of the device, responsible for executing instructions and performing calculations;

[0263] Memory: Used to store computer programs or instructions, as well as data during runtime;

[0264] Communication interface: used for exchanging data with other devices or satellites (e.g., receiving heartbeat signals, sending control commands);

[0265] Bus: The communication line connecting the above components;

[0266] When the processor executes a computer program or instructions stored in the memory, the electronic device can implement all the steps of the aforementioned method (implementations 1 to 6).

[0267] III. Computer-readable storage media:

[0268] A computer-readable storage medium having a computer program or instructions stored thereon; when these programs or instructions are executed by a processor, the steps of the aforementioned method can also be implemented.

[0269] Common storage media include:

[0270] Non-volatile storage devices: such as hard drives, SSDs, USB flash drives, optical discs, etc., whose data is not lost after power is turned off;

[0271] Solid-state storage devices: such as flash memory cards and embedded storage chips;

[0272] Other media capable of storing program instructions include magnetic tapes and disk arrays.

[0273] IV. Arrangement Device (System Module Division):

[0274] Corresponding to the aforementioned method, an orchestration device is also provided; the device is divided into the following seven modules according to their functions, each module corresponding to one or more steps in Implementation 1; these modules can be deployed on different hardware respectively, or they can be deployed together in the same device (such as a processor, controller or onboard computing platform).

[0275] 1. Status monitoring module:

[0276] Corresponding steps: Step 1.1 of Implementation Method 1 (Continuous monitoring).

[0277] Function: Responsible for continuously monitoring the ground control heartbeat, inter-satellite adjacency relationships, and the set of reachable nodes in the control domain. It is equivalent to the system's "eyes and ears," constantly collecting network status information.

[0278] 2. Split Confirmation Module:

[0279] Corresponding steps: Step 1.2 of Implementation Method 1 (splitting prediction and confirmation).

[0280] Function: Based on data provided by the status monitoring module, determine whether the split conditions (missing heartbeat + topology change) are met, and generate a partition epoch identifier upon confirmation. It acts as the system's "alarm," responsible for determining whether a network split has actually occurred.

[0281] 3. Snapshot Verification Module:

[0282] Corresponding steps: Steps 2.1 and 2.2 of Implementation Method 1 (State snapshot verification).

[0283] Function: Responsible for verifying and loading state snapshots. If a snapshot is corrupted, it triggers minimal autonomy mode (maintaining only existing instances and prohibiting the addition of low-priority instances). It acts as the system's "archiver," ensuring that the historical data used is complete and reliable.

[0284] 4. Restricted Autonomous Module:

[0285] Corresponding steps: Step 3 of Implementation Method 1 (Restricted Autonomy within the Partition).

[0286] Function: Calculates partition autonomy limits based on valid resource leases, critical business priorities, and risk budgets, and executes VNF deployments or migrations within those limits. It acts as a "temporary commander" for the system, adjusting services to a limited extent during partitions.

[0287] 5. Causal Log Module:

[0288] Corresponding step: Step 4 of Implementation Method 1 (Change Log Generation).

[0289] Function: Generates a change log for each mapping change during a split, including the partition epoch identifier, causal version vector, resource lease, and rollback point. It acts as the system's "bookkeeper," recording all modifications for later verification.

[0290] 6. Conflict Resolution Module:

[0291] Corresponding steps: Steps 5 and 6 of Implementation Method 1 (reconnection detection and conflict identification, conflict resolution).

[0292] Function: After reconnection, it exchanges snapshot summaries and change logs for each partition, identifies sequential updates and concurrent conflicts based on causal version vectors, and determines retention mappings according to lease validity, business priority, and risk budget consumption. If the first three are the same, retention mappings are selected according to the order of deterministic node identifiers. It acts as the system's "arbitrator," resolving configuration conflicts between different partitions.

[0293] 7. Progressive Recovery Module:

[0294] Corresponding step: Step 7 of Implementation Method 1 (Progressive Recovery and Unified Mapping).

[0295] Functionality: Establishes and retains target instances and paths corresponding to mappings, switches traffic in batches, performs health checks, and releases conflicting resources. If verification fails, it restores the original mapping based on the rollback point. It acts as the system's "surgeon," safely switching services from the old configuration to the new one.

[0296] V. Data transfer relationships between modules:

[0297] The above seven modules transfer data in the order of the method steps:

[0298] The outputs of the preceding modules (such as status data, split confirmation results, snapshots, autonomy quotas, change logs, conflict resolution results, etc.) serve as the inputs of the following modules, ultimately forming an orchestration result that can be deployed and executed and passes constraint review.

[0299] The name of a module is only used to describe its logical function and does not limit the number of specific software processes, hardware circuits, or deployment nodes. In other words, a single physical device can run multiple modules, and the same module can be distributed across multiple devices.

[0300] VI. Computer program products:

[0301] A computer program product is also provided, which includes a computer program or instructions; when these programs or instructions are executed by a processor, all the steps of the aforementioned methods (implementations 1 to 6) can be implemented. The computer program product typically exists in the form of an installation package or executable file, which users can download and install on their devices to run.

[0302] It should be noted that this invention improves the reliability and feasibility of orchestration in dynamic satellite networks by defining clear state formation, decision-making, security constraints, and transaction execution processes.

[0303] The above description of several specific embodiments further details the technical solution provided by the present invention in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above-described specific embodiments are not intended to limit the present invention. Any reasonable modifications and improvements to the present invention, reasonable combinations of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A satellite autonomous service orchestration method for inter-satellite network splitting and reconnection, characterized in that, The method includes the following steps: Monitor the ground control heartbeat, inter-satellite adjacency relationships, and the set of reachable nodes in the control domain; When the ground control heartbeat is continuously missing and the change in the set of reachable nodes in the control domain satisfies the splitting condition, the network split is confirmed, and a partition epoch identifier is generated to mark the current autonomous period. Verify and load a status snapshot containing the service function chain mapping version, node resource balance, resource lease and risk budget, and determine the partition autonomy quota based on the status snapshot; Within the autonomous quota of the partition, restricted virtual network functions are deployed or migrated for critical business operations, and a change log containing the partition epoch identifier, causal version vector, resource lease and rollback point is generated for each mapping change; After reconnection is detected, the snapshot summaries and change logs of each partition are exchanged, and updates and concurrent conflicts are identified according to the causal version vector. For the aforementioned concurrent conflicts, a retention mapping is determined based on lease validity, business priority, and risk budget consumption; Establish target instances and paths corresponding to the reserved mapping, switch traffic in batches, and release conflicting resources.

2. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection according to claim 1, characterized in that, The splitting conditions include: the number of consecutive missing ground control heartbeats reaches a first threshold, and the symmetric difference between the current control domain reachable node set and the previous control domain reachable node set reaches a second threshold.

3. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection according to claim 1, characterized in that, The partition epoch identifier includes the split event sequence number, the partition representative node identifier, and the time base, which are used to distinguish changes that occur during different autonomous periods.

4. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection according to claim 1, characterized in that, The state snapshot also includes the authorized quota, the list of key services, the resource reservation amount, the nearest feasible path, and the integrity verification value; when the integrity verification value indicates that the integrity verification of the state snapshot has failed, the partition enters the minimum autonomy mode, which maintains the existing instances and prohibits the addition of new low-priority instances.

5. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection according to claim 1, characterized in that, The process of determining the autonomous quota for the partition includes: The recommended amount is calculated based on the valid resource leases, key business priorities, and risk budgets in the aforementioned status snapshot; The suggested quota is compared with the authorized quota in the status snapshot and the value of the available idle resources in this partition after deducting the fault reservation amount, and the minimum value is taken as the final partition autonomy quota.

6. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection according to claim 1, characterized in that, The restricted deployment or migration of virtual network functions shall meet the following restrictions: operations shall be performed only on service function chains with a priority of not less than a preset level; the number of migrations of a single service function chain shall not exceed a preset number; and the total resources consumed by all operations shall not exceed the partition autonomy quota. Use only the resources authorized by valid resource leases that have not yet expired.

7. The on-board autonomous service orchestration method for inter-satellite network splitting and reconnection according to claim 1, characterized in that, The batch switching of traffic includes: the proportion of each batch of traffic does not exceed a preset proportion, and the health status, end-to-end latency and queue growth rate of the target instance are verified before the next batch of switching; if the verification fails, the subsequent switching is stopped and the original mapping is restored according to the rollback point.

8. An on-board autonomous service orchestration method and apparatus for inter-satellite network splitting and reconnection, characterized in that, The device includes the following modules: The status monitoring module is used to monitor the ground control heartbeat, inter-satellite adjacency relationships, and the set of reachable nodes in the control domain; The split confirmation module is used to confirm network splitting when the ground control heartbeat is continuously missing and the change in the set of reachable nodes in the control domain satisfies the splitting condition, and to generate a partition epoch identifier. The snapshot verification module is used to verify and load a status snapshot that includes the service function chain mapping version, node resource balance, resource lease and risk budget. The restricted autonomy module is used to determine the partition autonomy quota based on the state snapshot, and to perform virtual network function deployment or migration within the partition autonomy quota; The causal log module is used to generate change logs for mapping changes during splits, including the partition epoch identifier, causal version vector, resource lease, and rollback point. The conflict resolution module is used to exchange snapshot summaries and change logs of each partition after reconnection, identify sequential updates and concurrent conflicts based on the causal version vector, and determine the retention mapping according to lease validity, business priority and risk budget consumption. The progressive recovery module is used to establish target instances and paths corresponding to the retained mapping, switch traffic in batches, perform health verification and release conflicting resources, and restore the original mapping according to the rollback point when the verification fails.

9. The apparatus according to claim 8, characterized in that, The snapshot verification module is also used to trigger the minimum autonomy mode when the integrity verification of the state snapshot fails. The minimum autonomy mode maintains the existing instance and prohibits the addition of new low-priority instances. The conflict resolution module is also used to select the reserved mapping according to the deterministic node identifier order when the lease validity, business priority and risk budget consumption are all the same.

10. An electronic device, comprising: A processor and a memory, characterized in that the memory is used to store executable instructions of the processor, the processor being configured to execute, via executing the executable instructions, the on-board autonomous service orchestration method for inter-satellite network splitting and reconnection as described in any one of claims 1 to 7.