A multi-domain path transmission method under dynamic topology of a drone cluster

Abstract

The application discloses a kind of multi-domain inter-path transmission methods under the dynamic topology of unmanned aerial vehicle cluster, comprising: boundary candidate node periodically sends keep-alive message, maintains the soft state of adjacent relationship with adjacent boundary node, obtains inter-domain adjacent state information;In response to cross-domain communication request, trigger distributed detection, gather the determination result of each boundary candidate node based on target domain identifier query inter-domain adjacent state information, obtain candidate boundary node set;According to the information age and stability index of candidate path, the comprehensive quality score is calculated;High-quality candidate set is selected based on comprehensive quality score, and current active boundary node and standby boundary node are determined;According to the path quality change of active boundary node, the dynamic switching between active boundary and standby boundary is executed.The application improves the continuity and information timeliness of multi-domain inter-path transmission under the premise of controllable control overhead.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) communication, and particularly relates to a method for inter-domain path transmission under dynamic topology of UAV swarm. Background Technology

[0002] With the widespread application of drone swarms in emergency communication, collaborative reconnaissance, aerial inspection, and low-altitude logistics, drone nodes rely on self-organizing networks for state synchronization, task coordination, and data transmission. Unlike the relatively stable ground-based network environment, drone nodes are highly mobile, have short link lifetimes, and are susceptible to obstruction and interference, causing continuous fluctuations in network topology and even the splitting and merging of connected subsets. In multi-swarm collaboration or multi-entity joint operations, different drone swarms are typically controlled by different management entities, employing different routing strategies, identification management methods, and resource control rules. In engineering, a group of nodes sharing a management strategy and routing context is often abstracted as an "autonomous domain." Intra-domain communication can rely on a unified intra-domain routing mechanism, but when multiple domains coexist and interact in the airspace, cross-domain data transmission needs are inevitable, such as cross-domain situational information sharing, cross-domain command distribution, and collaborative control closed loops. To reduce inter-domain maintenance overhead, existing interconnection solutions have adopted the approach of concentrating the inter-domain control plane at the boundary nodes. This means that a small number of boundary nodes (gateways) are responsible for cross-domain interconnection, exchanging aggregated reachability information between boundary nodes. Ordinary nodes within the domain do not need to maintain the state across the entire domain; they only need to send cross-domain data to the boundary nodes to complete subsequent forwarding, thus keeping the overall control plane size manageable.

[0003] However, under the aforementioned "boundary compression" architecture, inter-domain information mainly remains at the boundary node level. Intra-domain nodes typically only know to "deliver data to the boundary node," but struggle to differentiate among multiple boundary candidates or proactively utilize backup paths to avoid inter-domain connectivity issues that are about to fail. Compression and aggregation inevitably reduce information granularity. In the highly dynamic environment of UAVs, characterized by rapid topology changes and short path lifetimes, inter-domain path states are more prone to becoming outdated, with "records still existing but reality having changed." This leads to data being sent to boundary nodes that no longer possess cross-domain capabilities, resulting in retries, detours, and handover oscillations. Existing solutions often rely on reachability judgments or static priorities as the basis for path selection, lacking an effective characterization of "whether the path information itself is still fresh." They struggle to answer the crucial question of "whether this inter-domain path information is still reliable," leading to frequent forwarding failures and latency jitter for time-sensitive business data (such as control commands and collaborative states) during cross-domain transmission. Therefore, how to enable nodes within a domain to complete boundary selection and backup switching based on a limited but effective path digest while maintaining controllable control overhead, and how to jointly judge the freshness and stability of path states to reduce forwarding failures and switching oscillations caused by outdated paths, has become an urgent problem to be solved in the current multi-domain networking of UAV swarms. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for inter-domain path transmission under dynamic topology of UAV swarms, comprising: Based on the keep-alive messages periodically sent by the boundary candidate nodes, maintain the soft state of the adjacency relationship between the boundary candidate nodes and the neighboring boundary nodes, and obtain the inter-domain adjacency state information. Based on the inter-domain adjacency status information, in response to cross-domain communication requests initiated by intra-domain communication nodes, distributed detection is triggered, and the judgment results of each boundary candidate node querying the inter-domain adjacency status information based on the target domain identifier are aggregated to obtain a set of candidate boundary nodes. A comprehensive quality score is calculated based on the information age and stability index of the inter-domain path corresponding to each candidate boundary node in the candidate boundary node set. Based on the comprehensive quality score, a preset number of candidate boundary nodes with the highest quality are selected from the candidate boundary node set to form a high-quality candidate set, and the one with the highest quality is determined as the current active boundary node, while the remaining nodes are determined as backup boundary nodes. Based on the changes in path quality of the currently active boundary node, a dynamic switch is performed between the active boundary node and the backup boundary node.

[0005] Preferably, the process of obtaining the inter-domain adjacency state information includes: The candidate boundary node periodically sends boundary keep-alive messages to neighboring boundary nodes and receives keep-alive response messages from the other end. Based on the received results, maintain a neighborhood boundary adjacency table. Each entry in the neighborhood boundary adjacency table includes the identifier of the peer boundary node, the identifier of the autonomous system to which the peer belongs, the time of the most recent successful message reception, the consecutive keep-alive loss count, the recent keep-alive success rate, and the current adjacency status. Based on the continuous keep-alive loss count and the preset degradation threshold and failure threshold, the current adjacency state is controlled to transition between the active state, the degraded state and the failure state.

[0006] Preferably, the condition for the current adjacency state to transition from the active state to the degraded state is that the continuous keep-alive loss count reaches the degradation threshold but does not reach the failure threshold; The condition for the current adjacency state to transition from the degraded state or the active state to the failed state is that the continuous keep-alive loss count reaches the failure threshold; When the current adjacency state is in the degenerate state and the corresponding boundary candidate node is the currently active boundary node, the keep-alive detection frequency of the corresponding boundary candidate node is increased.

[0007] Preferably, the process of triggering distributed detection and obtaining the candidate boundary node set includes: The cross-domain communication request is propagated within the domain in a restricted manner through the intra-domain communication nodes, so that each boundary candidate node receives the request. Each boundary candidate node queries the locally maintained inter-domain adjacency status information based on the target domain identifier in the request; If a valid adjacency relationship matching the target domain identifier is found, the detection result can be reached by a single-point backhaul along the reverse path within the domain from the boundary candidate node. All reachability detection results returned are aggregated to form the candidate boundary node set.

[0008] Preferably, the process of calculating the comprehensive quality score includes: The information age is calculated based on the difference between the most recent successful update time of the inter-domain path and the current time. The age of the information is normalized and indexed to obtain a timeliness score. The stability score is calculated based on the percentage of continuous contact time of the inter-domain paths within the observation window, the keep-alive success rate, and the penalty for changes in adjacency state. The timeliness score and the stability score are weighted and summed to obtain the comprehensive quality score.

[0009] Preferably, the calculation process of the stability score includes: Obtain the percentage of continuous contact time, keep-alive success rate, and adjacency state change penalty for the inter-domain paths within the observation window; The stability score is obtained by weighting and summing the continuous contact time percentage, keep-alive success rate and adjacency state change penalty item by assigning preset weights to each of them. The adjacency state change penalty term is used to characterize the negative impact of the number of state changes of the inter-domain path within the observation window on stability.

[0010] Preferably, the process of obtaining a high-quality candidate set includes: Sort each candidate boundary node in the candidate boundary node set from high to low according to the corresponding comprehensive quality score; The high-quality candidate set is formed by selecting the first preset number of candidate boundary nodes after sorting. The candidate boundary node with the highest comprehensive quality score in the high-quality candidate set is determined as the current active boundary node.

[0011] Preferably, the process of performing dynamic switching includes: Monitor the path quality of the currently active boundary node; When the overall quality score of the current active boundary node decreases and the difference between the overall quality score of the new best candidate boundary node and the overall quality score of the current active boundary node exceeds a preset handover hold threshold, an active boundary handover is performed. When the current active boundary node fails, a valid node is selected from the backup boundary nodes in descending order of the comprehensive quality score to take over the active boundary responsibilities.

[0012] Preferably, the preset switching hold threshold is a positive value; When the difference between the overall quality score of the new best candidate boundary node and the overall quality score of the current active boundary node is greater than the preset handover hold threshold, the active boundary handover is triggered. When the current active boundary node fails, priority is given to trying to take over from the backup boundary node with the second highest comprehensive quality score in the high-quality candidate set. If the backup boundary node is unavailable, subsequent backup boundary nodes are tried in order of sorting.

[0013] Preferably, the process of performing the dynamic switching further includes: When all candidate boundary nodes in the high-quality candidate set fail, the distributed detection is retried to update the candidate boundary node set. The filtering and switching operations are re-executed based on the updated set of candidate boundary nodes.

[0014] Compared with the prior art, the present invention has the following advantages and technical effects: (1) In view of the communication establishment delay caused by the need to temporarily perform full boundary detection in the existing boundary compression architecture, the present invention designs a pre-maintenance mechanism for boundary keep-alive messages, which decouples the discovery of boundary adjacency state from the triggering of cross-domain requests, so that when a node in the domain initiates cross-domain communication, it can directly obtain the reachability judgment based on the maintained adjacency soft state, which significantly reduces the establishment delay of cross-domain communication and improves the response speed of time-sensitive services.

[0015] (2) In view of the problem that the compressed inter-domain information in the existing scheme lacks freshness characterization and intra-domain nodes are prone to make incorrect forwarding decisions based on outdated boundary information, the present invention introduces a joint evaluation model based on information age and topology stability. On the basis of judging the reachability of candidate paths, the reliability of path information is further comprehensively quantified from the time dimension and historical stability dimension, which effectively avoids cross-domain forwarding errors caused by using invalid or expired boundary information and improves the accuracy of inter-domain path selection.

[0016] (3) In response to the problem that frequent switching of active boundaries caused by short-term topology fluctuations in existing schemes, which in turn leads to path oscillations, this invention introduces a high-quality candidate set screening mechanism and a switching hold threshold. By retaining multiple backup boundary nodes and setting a quality improvement margin in the switching conditions, the boundary oscillations caused by short-term signal fluctuations are suppressed while ensuring path quality, and the smooth adjustment of active boundaries is achieved. At the same time, the existence of backup sets allows the system to quickly complete the local replacement when the main path fails, shortening the fault recovery time and improving the continuity and stability of inter-domain transmission as a whole. Attached Figure Description

[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the maintenance and adjacency keeping of boundary candidate nodes in an embodiment of the present invention. Figure 3 This is a schematic diagram of distributed detection and result feedback triggered by Interest in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the dynamic selection and switching of boundary nodes according to an embodiment of the present invention; Figure 5 This is a graph showing the experimental results comparing the success rate of inter-domain communication under different node movement speeds in an embodiment of the present invention. Figure 6This is a graph showing the experimental results comparing path maintenance overhead in different scenarios according to embodiments of the present invention; Figure 7 This is a graph showing the experimental results comparing the number of activity boundary switching times with the average recovery time in an embodiment of the present invention. Detailed Implementation

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0020] like Figure 1 As shown, this embodiment provides a method for inter-domain path transmission under a dynamic topology of a drone swarm, including: Based on the keep-alive messages periodically sent by the candidate boundary nodes, maintain the soft state of the adjacency relationship between the candidate boundary nodes and the neighboring boundary nodes, and obtain the inter-domain adjacency state information. Based on the inter-domain adjacency status information, in response to cross-domain communication requests initiated by intra-domain communication nodes, distributed detection is triggered, and the judgment results of each boundary candidate node based on the target domain identifier querying the inter-domain adjacency status information are aggregated to obtain a set of candidate boundary nodes; Calculate the overall quality score based on the information age and stability index of the inter-domain path corresponding to each candidate boundary node in the candidate boundary node set; Based on the comprehensive quality score, a preset number of candidate boundary nodes with the highest quality are selected from the candidate boundary node set to form a high-quality candidate set. The one with the highest quality is determined as the current active boundary node, and the remaining nodes are determined as backup boundary nodes. Based on changes in the path quality of the current active boundary node, a dynamic switch is performed between the active boundary node and the backup boundary node.

[0021] In this embodiment, nodes listen to neighbor packets and maintain a neighbor table. When a neighbor from another domain is detected, it is identified as a candidate boundary node. The candidate boundary node performs reachability detection on the external domain, constructing inter-domain path entries containing the destination domain identifier, next-hop boundary node, most recent acknowledgment timestamp, and lifetime. These entries are distributed within the domain through periodic announcements and maintained and cleared using a soft-state approach. For each inter-domain path entry, the Information Age (AoI) is calculated to characterize the freshness of the path information, and combined with inter-domain stability indicators to form a comprehensive quality score. Based on the comprehensive quality score, no more than K active boundary nodes are selected from the candidate boundary node set to form a primary / backup set for cross-domain forwarding. When the primary path fails or the score drops below a threshold, the node switches to the backup set, and hysteresis and cooling-off time are used to suppress frequent switching oscillations. This invention improves the continuity and timeliness of inter-domain transmission while keeping overhead controllable, and is suitable for multi-domain collaborative communication scenarios in UAV swarms.

[0022] Furthermore, this embodiment specifically relates to a method for multi-domain path transmission and path management in a dynamic topology environment of UAV swarms. More specifically, it relates to a communication mechanism for multi-domain forwarding and fast switching under the conditions of multiple autonomous domains coexisting, dynamic changes in domain boundary nodes, and time-varying inter-domain links. This mechanism is achieved through inter-domain path discovery, soft state maintenance, timeliness assessment based on information age, stability modeling, and dynamic selection of boundary nodes. It can be applied to cross-domain data transmission scenarios in UAV swarm collaborative tasks.

[0023] Furthermore, the process of obtaining inter-domain adjacency state information includes: The candidate boundary node periodically sends boundary keep-alive messages to neighboring boundary nodes and receives keep-alive response messages from the peer. Based on the received results, maintain the neighborhood boundary adjacency table. Each entry in the neighborhood boundary adjacency table contains the identifier of the peer boundary node, the identifier of the autonomous system to which the peer belongs, the time of the most recent successful message reception, the consecutive keep-alive loss count, the recent keep-alive success rate, and the current adjacency status. Based on the continuous keep-alive loss count and preset degradation and failure thresholds, the current adjacent state is controlled to transition between active, degraded, and failed states.

[0024] Furthermore, the condition for the current adjacency state to transition from the active state to the degraded state is that the continuous keep-alive loss count reaches the degradation threshold but does not reach the failure threshold; The condition for the current adjacency state to transition from a degraded state or a valid state to a failed state is that the continuous keep-alive loss count reaches the failure threshold. When the current adjacency state is in a degenerate state and the corresponding boundary candidate node is the current active boundary node, increase the keep-alive detection frequency of the corresponding boundary candidate node.

[0025] Furthermore, in drone swarm networks, multiple drones performing different tasks form relatively independent autonomous domains (ADAs). These ADAs need to continuously exchange state information, perception results, and task instructions. Due to the high-speed maneuverability of drone nodes, inter-domain adjacency relationships are established and disappear quickly, making traditional static routing protocols unsuitable for this scenario. This embodiment proposes an inter-domain path management method that uses a lightweight keep-alive mechanism to pre-maintain the soft state of boundary adjacencies, combines information age and topology stability to jointly evaluate the quality of candidate paths, and uses top-k filtering and switching thresholds to dynamically select active boundaries.

[0026] In a dynamic topology environment, the identity of a boundary candidate node is not statically given, but rather determined by whether the node is continuously at the domain edge and continuously in contact with external domains. This invention uses a soft-state approach to maintain the identity of boundary candidate nodes. Any node that can stably contact its neighboring domains and has the potential to assume out-of-domain responsibilities for its own domain can be considered a boundary candidate node. Once its external contact relationship disappears within a certain period of time, its boundary candidate status is downgraded or revoked.

[0027] like Figure 2 As shown, to maintain the aforementioned soft state, candidate boundary nodes periodically send Boundary Hello messages to neighboring boundary nodes to maintain inter-domain adjacency relationships. The Boundary Hello mechanism primarily answers two questions: first, whether the current node is still in contact with a certain neighboring domain; and second, whether the peer is still a valid neighboring boundary node.

[0028] A Boundary Hello message must contain at least the following fields: sending node identifier, domain identifier, boundary role, message sequence number, sending timestamp, and keep-alive validity period. In addition, the message may carry a reply control field to indicate whether a reply from the peer is required. When Hello is only used to unidirectionally refresh the existence of the local boundary, a reply is not required; when Hello is used to explicitly renew a certain adjacency relationship, a HelloReply response from the peer is required. Therefore, Hello can serve both periodic identity announcement functions and lightweight adjacency confirmation functions.

[0029] Regarding the transmission frequency, the boundary Hello should be adaptively adjusted based on state changes. For stable neighboring relationships, a lower frequency should be used to maintain them; when a relationship experiences jitter, loses several consecutive responses, or recently assumes the responsibility of the active boundary, the transmission frequency should be temporarily increased to more quickly determine whether it has degraded or failed.

[0030] Based on the above mechanism, each candidate boundary node maintains a neighborhood boundary adjacency table. Any entry in this table can be represented as: Among them, NID i DID represents the identifier of the peer boundary node. i Indicates the autonomous system identifier of the peer, t i last This indicates the time when the "Hello" or "Reply" message was last successfully received, m i Represents the continuous keep-alive loss count, ρ i This indicates the recent survival success rate, s i This indicates the current adjacency state.

[0031] Let m be the consecutive loss count of a certain candidate boundary relationship. i The degradation threshold is θ d The failure threshold is θ f , where θ f >θ d The path state transition rule is as follows: It should be noted that path state degradation and failure are not directly triggered by a single Hello loss, but are determined by the cumulative effect of consecutive mismatches. This is because links in a drone swarm network are highly susceptible to occlusion, maneuvering, and short-term interference. If the network is too sensitive to a single mismatch, the path will be frequently canceled and rebuilt during short-term jitter, which will reduce the stability of the mechanism.

[0032] When an adjacency enters a degraded state and the node is currently responsible for the activity boundary, the maintenance module temporarily increases its keep-alive detection frequency to more quickly determine its subsequent trend. If the degraded state has not recovered within a longer observation window, the adjacency is marked as invalid, the corresponding candidate path is withdrawn, and a distributed detection and candidate boundary reconstruction process is immediately triggered.

[0033] Furthermore, the process of triggering distributed detection and obtaining a set of candidate boundary nodes includes: The cross-domain communication request is propagated within the domain in a restricted manner by the intra-domain communication node, so that each boundary candidate node can receive the request. Each boundary candidate node queries the locally maintained inter-domain adjacency status information based on the target domain identifier in the request; If a valid adjacency relationship matching the target domain identifier is found, the detection result can be reached by a single point backhaul along the reverse path within the domain from the boundary candidate node. All reachability detection results returned are aggregated to form a set of candidate boundary nodes.

[0034] Furthermore, the process of calculating the overall quality score includes: Calculate the information age based on the difference between the most recent successful update time of the inter-domain path and the current time. A timeliness score is obtained by normalizing the age of information using an index mapping. The stability score is calculated based on the percentage of continuous contact time between inter-domain paths within the observation window, the survival success rate, and the penalty for changes in adjacency state. The timeliness score and the stability score are weighted and summed to obtain the overall quality score.

[0035] Furthermore, when a communication node within a domain needs to initiate cross-domain communication to a target domain, it does not directly know which candidate boundary nodes within its own domain currently possess the capability to reach the target domain. Therefore, a detection process targeting candidate boundary nodes is required to quickly converge and determine "who can reach the target domain" within the current domain.

[0036] like Figure 3 As shown, the distributed detection process in this embodiment can be summarized as "request triggers query, query depends on pre-maintained state, and result is sent back and aggregated at a single point". Specifically, after a communication node within the domain initiates a cross-domain communication request Interest for the first time, the request is diffused within the domain to each boundary candidate node in a limited manner; after each boundary candidate node receives the Interest, it does not continue to diffuse within the domain, nor does it send a new Hello response to the neighboring domain by default, but instead directly queries the local neighboring domain boundary adjacency table based on the target domain identifier in the request.

[0037] Let the target domain requested by the communication node within this domain be DID. d Then any candidate node b on the boundary i The local decision function is: When Γ(b) i DID d When )=1, the candidate boundary node considers itself as a feasible candidate boundary and sends back the detection result along the reverse path within the domain; otherwise, it does not send back the result, or only sends back the unreachable marker. The returned detection result includes at least: the identifier of this candidate boundary node, the identifier of the target domain, the identifier of the matched neighboring boundary node, the time of the most recent effective adjacency with the target domain, the current adjacency status, and necessary local quality parameters.

[0038] After the source node aggregates all returned results, it constructs a target domain-oriented D. j The set of candidate boundary nodes: If the set is empty, the cross-origin request is cached or awaits the next round of detection; if the set is not empty, proceed to step three for path quality assessment.

[0039] If the set is empty, the cross-origin request is cached or awaits the next round of detection; if the set is not empty, proceed to step three for path quality assessment.

[0040] Furthermore, the calculation process for the stability score includes: Obtain the percentage of continuous contact time, keep-alive success rate, and adjacency state change penalty for inter-domain paths within the observation window; After assigning preset weights to the percentage of continuous contact time, the survival success rate, and the penalty for changes in adjacency state, the stability score is obtained by weighted summation. Among them, the adjacency state change penalty term is used to characterize the negative impact of the number of state changes of inter-domain paths within the observation window on stability.

[0041] Furthermore, the method for comprehensive quality evaluation of candidate paths in this embodiment includes: Let t be the time of the most recent successful update of the candidate path. u The current time is t, and the information age is defined as: Timeliness scores are mapped to normalized exponential form: Where τ is the time-related decay parameter; the stability score is defined as: Where D i R represents the percentage of continuous contact time within the observation window. i For Boundary Hello to successfully maintain the ratio, P i The adjacency state change penalty term, α, β, γ > 0 are weight parameters; the overall quality score is defined as: Furthermore, the process of obtaining a high-quality candidate set includes: Sort the candidate boundary nodes in the candidate boundary node set from high to low according to their corresponding comprehensive quality scores; Select the first preset number of candidate boundary nodes after sorting to form a high-quality candidate set; The candidate boundary node with the highest overall quality score in the high-quality candidate set is determined as the current active boundary node.

[0042] Furthermore, after obtaining the set of candidate boundary nodes and completing the path quality assessment, this embodiment introduces a top-k screening mechanism to determine the active boundary nodes and backup boundary nodes.

[0043] like Figure 4 As shown, the specific method is as follows.

[0044] First, for set B jThe candidate nodes are sorted from highest to lowest overall quality to obtain an ordered candidate sequence: Based on this, the top k nodes with the highest quality are selected to form a high-quality candidate boundary set: Here, k is a system preset parameter used to control the number of high-quality nodes that ultimately participate in the activity boundary competition. When the number of candidate nodes is less than k, all candidate nodes are actually selected.

[0045] In this embodiment, the top-k mechanism does not mean simultaneously enabling k active boundary nodes for concurrent forwarding. Instead, it selects the k highest-quality nodes from all candidate boundary nodes to form a high-quality candidate set. The node with the highest overall quality is then determined as the current active boundary node. The remaining nodes serve as backup boundary nodes, participating in subsequent replacements when the current active boundary degrades or fails. If multiple candidate nodes have the same overall quality, the one with the most recent state update time is selected first; if they still cannot be distinguished, the one with the smaller boundary node identifier is further selected.

[0046] Furthermore, the process of performing dynamic switching includes: Monitor the path quality of the current active boundary nodes; When the overall quality score of the current active boundary node decreases and the difference between the overall quality score of the new best candidate boundary node and the overall quality score of the current active boundary node exceeds the preset handover hold threshold, the active boundary handover is performed. When the current activity boundary node fails, valid nodes are selected from the backup boundary nodes in descending order of their comprehensive quality scores to take over the activity boundary responsibilities.

[0047] Furthermore, the preset handover hold threshold is set to a positive value; When the difference between the overall quality score of the new best candidate boundary node and the overall quality score of the current active boundary node is greater than the preset handover hold threshold, the active boundary handover is triggered. When the current active boundary node fails, priority is given to trying to replace it from the backup boundary node with the second highest overall quality score in the high-quality candidate set. If the backup boundary node is unavailable, the subsequent backup boundary nodes are tried in order of sorting.

[0048] Furthermore, the process of performing dynamic switching also includes: When all candidate boundary nodes in the high-quality candidate set fail, the distributed detection is retried to update the candidate boundary node set. The filtering and switching operations are re-executed based on the updated set of candidate boundary nodes.

[0049] Furthermore, in a dynamic topology environment, immediately switching the active boundary whenever a higher-quality node appears can easily cause boundary oscillations. Therefore, this embodiment introduces a switching hold threshold Δ. Let the current active boundary quality be Qc, and the quality of the new best candidate node be Qn. The system will only perform an active boundary switch if the following conditions are met: Otherwise, the current activity boundaries will remain unchanged.

[0050] The dynamic adjustment of the current active boundary node follows these four rules: Rule 1: If the current activity boundary state is valid, and there are no new candidate nodes significantly better than it (i.e., Q...) n -Q c If the boundary value is less than or equal to Δ, then the current activity boundary remains unchanged. Rule 2: If a new candidate node exists whose quality improvement exceeds the threshold Δ, then it is switched to the new activity boundary; Rule 3: If the current active boundary fails, the next valid node is promoted from the top-k spare set in sorted order. For example, when b(1) fails, b(2) is tried first; if b(2) is also unavailable, b(3) is tried, and so on. Rule 4: If all candidate nodes in the top-k set fail, the Interest-triggered distributed detection process in step 2 is retried to rebuild the candidate boundary set.

[0051] Through the above mechanism, the system can smoothly adjust the activity boundary when path quality degrades, boundary fails, and neighborhood is reconstructed, maintaining the continuity of inter-domain communication while minimizing unnecessary switching.

[0052] In the above implementation, this embodiment verifies a multi-domain UAV swarm scenario built on the OMNeT++ simulation platform. The simulation platform contains 3 to 6 autonomous domains, with 10 to 30 nodes per domain. Node movement speeds are divided into low, medium, and high speeds. The simulation also includes parameters such as the Hello cycle, the number of candidate paths retained (k), and the failure threshold (θ). f The switching hold threshold Δ and the service transmission rate are all adjustable parameters.

[0053] Figure 5The results show a comparison of the inter-domain communication success rates of the method in this embodiment and the single-path baseline scheme under different node movement speeds. It can be found that the communication success rate of all schemes decreases as the node movement speed increases. However, this embodiment, due to the introduction of candidate path set maintenance, dynamic selection of activity boundaries, and anomaly recovery mechanism, can still maintain a high communication success rate in medium-to-high dynamic scenarios, demonstrating the advantages of this embodiment in ensuring boundary continuity.

[0054] Figure 6 The results show a comparison of path maintenance overhead between the proposed method and the baseline scheme under different scenarios. Under low service load, the main overhead comes from boundary adjacency keep-alive; as service load increases, the overhead of cross-domain requests and candidate boundary coordination messages also increases, but the overall control plane burden remains within an acceptable range. This is mainly because this embodiment restricts the control plane primarily to the boundary candidate nodes and active boundary nodes, eliminating the need for ordinary intra-domain nodes to maintain cross-domain states for extended periods.

[0055] Figure 7 The results demonstrate the comparison between the number of activity boundary handovers and the average recovery time under different k values ​​and handover thresholds Δ in a high-dynamic scenario. The results show that introducing the handover holding threshold Δ significantly reduces the number of activity boundary handovers, indicating that this mechanism can effectively suppress unnecessary handovers caused by short-term jitter. Simultaneously, due to the existence of the top-k backup set, when the main path truly fails, the system can usually complete a rapid takeover within the local candidate set, significantly shortening the average recovery time.

[0056] The Boundary Hello pre-maintenance mechanism designed in this embodiment decouples the discovery of boundary adjacency state from the triggering of cross-domain requests, avoiding the re-execution of full boundary detection every time an Interest arrives, and significantly reducing the latency of cross-domain communication establishment.

[0057] The joint evaluation model based on AoI and topology stability proposed in this embodiment simultaneously characterizes the timeliness and historical stability of candidate path states. Compared with a single reachability judgment, it can more accurately reflect the continuous availability of boundary nodes and effectively avoid cross-domain forwarding errors caused by using outdated boundary information.

[0058] This embodiment introduces a top-k filtering mechanism and a switching hold threshold Δ, which ensures the lower limit of path quality while suppressing boundary oscillations caused by short-term fluctuations, realizes smooth adjustment of active boundary nodes, and improves the continuity and stability of path transmission between multiple domains.

[0059] As an additional implementation, this embodiment also provides an inter-domain path management system for multi-domain networks of unmanned aerial vehicle (UAV) swarms. The system is used to implement the method and includes: a keep-alive module, a detection module, an evaluation module, a selection module, and a switching module.

[0060] The keep-alive module is used to maintain the identity of candidate nodes at the boundary and keep them alive in terms of adjacency relationships. This module periodically sends Boundary Hello messages, maintains the neighborhood boundary adjacency table, and updates the adjacency list based on the consecutive loss count m. i Degradation threshold θ d Failure threshold θ f The driving adjacency state transitions between the three states of valid, degraded, and invalid.

[0061] The detection module is used to perform distributed detection triggered by the interest and return the results, and to construct a candidate boundary node set B. j Upon receiving a cross-domain communication request, this module will diffuse Interest within its local domain in a restricted manner, collecting candidate nodes at each boundary and executing the decision function Γ(b) based on their local adjacency lists. i DID d The results of the single-point return are then aggregated to form a candidate set.

[0062] The evaluation module is used to jointly evaluate the timeliness and stability of candidate paths and calculate the comprehensive quality score Q. i =w1F i +w2S i This module bases each candidate node on its AoI, keep-alive success rate Ri, continuous contact rate Di, and adjacency state change penalty term P. i Calculate F i With S i The scores are then weighted and combined to obtain a comprehensive quality score.

[0063] The selection module is used to perform top-k filtering on the candidate set B. j Sort by overall quality, and select the top k nodes to form a high-quality candidate set B. j (k), and the node with the highest overall quality is determined as the activity boundary node b. j The remaining nodes serve as backup boundary nodes.

[0064] The switching module is used to dynamically monitor changes in path quality and determine whether to perform an active boundary switch based on the switching retention threshold Δ. When the active boundary fails, it is promoted sequentially from the top-k backup set. When the entire backup set fails, the detection module is re-triggered to perform candidate set reconstruction to achieve anomaly recovery.

[0065] In this embodiment, the core calculation process of each module corresponds exactly to the calculation formula of each step in Embodiment 1, and will not be repeated here.

[0066] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for inter-domain path transmission under dynamic topology of UAV swarm, characterized in that, include: Based on the keep-alive messages periodically sent by the boundary candidate nodes, maintain the soft state of the adjacency relationship between the boundary candidate nodes and the neighboring boundary nodes, and obtain the inter-domain adjacency state information. Based on the inter-domain adjacency status information, in response to cross-domain communication requests initiated by intra-domain communication nodes, distributed detection is triggered, and the judgment results of each boundary candidate node querying the inter-domain adjacency status information based on the target domain identifier are aggregated to obtain a set of candidate boundary nodes. A comprehensive quality score is calculated based on the information age and stability index of the inter-domain path corresponding to each candidate boundary node in the candidate boundary node set. Based on the comprehensive quality score, a preset number of candidate boundary nodes with the highest quality are selected from the candidate boundary node set to form a high-quality candidate set, and the one with the highest quality is determined as the current active boundary node, while the remaining nodes are determined as backup boundary nodes. Based on the changes in path quality of the currently active boundary node, a dynamic switch is performed between the active boundary node and the backup boundary node.

2. The method according to claim 1, characterized in that, The process of obtaining the inter-domain adjacency state information includes: The candidate boundary node periodically sends boundary keep-alive messages to neighboring boundary nodes and receives keep-alive response messages from the other end. Based on the received results, maintain a neighborhood boundary adjacency table. Each entry in the neighborhood boundary adjacency table includes the identifier of the peer boundary node, the identifier of the autonomous system to which the peer belongs, the time of the most recent successful message reception, the consecutive keep-alive loss count, the recent keep-alive success rate, and the current adjacency status. Based on the continuous keep-alive loss count and the preset degradation threshold and failure threshold, the current adjacency state is controlled to transition between the active state, the degraded state and the failure state.

3. The method according to claim 2, characterized in that, The condition for the current adjacency state to transition from the active state to the degraded state is that the continuous keep-alive loss count reaches the degradation threshold but does not reach the failure threshold; The condition for the current adjacency state to transition from the degraded state or the active state to the failed state is that the continuous keep-alive loss count reaches the failure threshold; When the current adjacency state is in the degenerate state and the corresponding boundary candidate node is the currently active boundary node, the keep-alive detection frequency of the corresponding boundary candidate node is increased.

4. The method according to claim 1, characterized in that, The process of triggering distributed detection and obtaining the candidate boundary node set includes: The cross-domain communication request is propagated within the domain in a restricted manner through the intra-domain communication nodes, so that each boundary candidate node receives the request. Each boundary candidate node queries the locally maintained inter-domain adjacency status information based on the target domain identifier in the request; If a valid adjacency relationship matching the target domain identifier is found, the detection result can be reached by a single-point backhaul along the reverse path within the domain from the boundary candidate node. All reachability detection results returned are aggregated to form the candidate boundary node set.

5. The method according to claim 1, characterized in that, The process of calculating the comprehensive quality score includes: The information age is calculated based on the difference between the most recent successful update time of the inter-domain path and the current time. The age of the information is normalized and indexed to obtain a timeliness score. The stability score is calculated based on the percentage of continuous contact time of the inter-domain paths within the observation window, the keep-alive success rate, and the penalty for changes in adjacency state. The timeliness score and the stability score are weighted and summed to obtain the comprehensive quality score.

6. The method according to claim 5, characterized in that, The calculation process for the stability score includes: Obtain the percentage of continuous contact time, keep-alive success rate, and adjacency state change penalty for the inter-domain paths within the observation window; The stability score is obtained by weighting and summing the continuous contact time percentage, keep-alive success rate and adjacency state change penalty item by assigning preset weights to each of them. The adjacency state change penalty term is used to characterize the negative impact of the number of state changes of the inter-domain path within the observation window on stability.

7. The method according to claim 1, characterized in that, The process of obtaining a high-quality candidate set includes: Sort each candidate boundary node in the candidate boundary node set from high to low according to the corresponding comprehensive quality score; The high-quality candidate set is formed by selecting the first preset number of candidate boundary nodes after sorting. The candidate boundary node with the highest comprehensive quality score in the high-quality candidate set is determined as the current active boundary node.

8. The method according to claim 1, characterized in that, The process of performing dynamic switching includes: Monitor the path quality of the currently active boundary node; When the overall quality score of the current active boundary node decreases and the difference between the overall quality score of the new best candidate boundary node and the overall quality score of the current active boundary node exceeds a preset handover hold threshold, an active boundary handover is performed. When the current active boundary node fails, a valid node is selected from the backup boundary nodes in descending order of the comprehensive quality score to take over the active boundary responsibilities.

9. The method according to claim 8, characterized in that, The preset switching hold threshold is a positive value; When the difference between the overall quality score of the new best candidate boundary node and the overall quality score of the current active boundary node is greater than the preset handover hold threshold, the active boundary handover is triggered. When the current active boundary node fails, priority is given to trying to take over from the backup boundary node with the second highest comprehensive quality score in the high-quality candidate set. If the backup boundary node is unavailable, subsequent backup boundary nodes are tried in order of sorting.

10. The method according to claim 1, characterized in that, The process of performing the dynamic switching also includes: When all candidate boundary nodes in the high-quality candidate set fail, the distributed detection is retried to update the candidate boundary node set. The filtering and switching operations are re-executed based on the updated set of candidate boundary nodes.

Images