A large-scale uanet clustering networking method based on information entropy dynamic weighting and grace period model

Abstract

The application provides a large-scale UANET clustering networking method based on information entropy dynamic weighting and grace period model, characterized in that the implementation steps are strictly carried out according to system control logic, including the following steps: step 1: initializing unmanned aerial vehicle network and cold start networking based on static weight; step 2: super cluster head calculates information entropy based on state feature matrix and issues dynamic weight set; step 3: steady-state clustering reselection and master-backup binding based on dynamic weight; step 4: multi-dimensional steady-state maintenance and seamless role switching based on grace period. The application is a brand-new large-scale clustering networking and dynamic maintenance method, which can realize objective intelligent networking of high dynamic node cluster and seamless smooth evolution of topology under the premise of extremely low control overhead.

Description

Technical Field

[0001] This invention belongs to the field of UANET topology management and communication networking technology, specifically involving a large-scale UANET clustering networking method based on dynamic weighting of information entropy and grace period model. Background Technology

[0002] With the rapid development of unmanned systems technology, drones, due to their advantages of maneuverability, low cost, rapid deployment, and versatility, are being rapidly applied in many fields such as emergency communication, disaster relief, power line inspection, close-range battlefield reconnaissance, and military strikes. However, a single drone is limited by its own energy, communication capabilities, and payload, making it difficult to independently complete complex tasks. Multi-drone collaborative operations have become an inevitable trend. Unmanned Aerial Vehicle Ad Hoc Network (UANET) has thus become a current research hotspot.

[0003] Compared to traditional ground mobile ad hoc networks (MANET / Ad Hoc), UANET has extremely unique network characteristics, and its network environment is more severe and complex: First, sudden node disappearance is common. UAVs often leave the network without warning due to reasons such as low battery return, obstacle avoidance maneuvers, task switching, or wind aerodynamic instability. This is the core reason for the high instability of the network topology. Second, communication energy consumption accounts for a high proportion and the energy model is complex. The transient transmission power consumption of UAV communication modules is large. Nodes that overload forwarding tasks will quickly run out of power, become "hot spots," and fail prematurely. Third, the airspace environment has a significant impact on propagation characteristics. The link is often in a mixed state of free space and two-path fading. Affected by wind disturbances and changes in aircraft attitude, the link quality exhibits strong time-varying characteristics and large random fluctuations.

[0004] In recent years, with the evolution of UAV control technology and the reduction in deployment costs, the scale of UAV ad hoc networks has been expanding from small formations of a dozen or so UAVs to ultra-large clusters of two hundred or even thousands. In large-scale scenarios exceeding one hundred UAVs, the topological connectivity between nodes grows exponentially, further amplifying the inherent high dynamism and concentrated energy consumption inherent in UANET. Faced with such complex scenarios, the physical limitations of traditional networking mechanisms are fully exposed, mainly encountering the following three major technical bottlenecks: First, there's the issue of signaling overhead explosion and broadcast storms. Traditional decentralized, flat networking protocols, such as AODV and OLSR, heavily rely on network-wide flooding to maintain routing and topology. When the number of network nodes exceeds a certain threshold, the control signaling required for topology detection and maintenance consumes a significant amount of wireless bandwidth. During highly dynamic flight scenarios that lead to frequent topology reconfigurations, this can easily trigger widespread "broadcast storms," ​​causing severe queuing delays and packet loss for valid service data.

[0005] Second, subjective weighting fails in multidimensional heterogeneous states. Faced with the high heterogeneity of massive nodes in terms of remaining energy, communication load, and link quality, existing clustering algorithms, such as the traditional WCA algorithm, mostly adopt static empirical weighting formulas. This fixed weight allocation cannot keenly perceive the dynamic changes of the network at different mission stages, such as the connectivity bottleneck at the beginning of takeoff and the power bottleneck at the end of the mission. It is difficult to coordinate the overall state and easily leads to local resource misallocation in the network.

[0006] Third, topology oscillations caused by single-point failures. Traffic in large-scale networks tends to converge towards backbone nodes. If a cluster head fails due to sudden large-scale traffic congestion or energy depletion, the original cross-cluster routing path is instantly physically severed. This hard disconnection immediately triggers a large number of surrounding member-level nodes to initiate emergency paging and local reconstruction, thereby causing continuous topology oscillations and ineffective energy consumption.

[0007] In summary, existing technologies suffer from systemic flaws in three dimensions—topology hierarchy, node election criteria, and steady-state maintenance mechanisms—when dealing with large-scale UAV swarm networking. These flaws easily lead to a vicious cycle of control signaling congestion, network-wide resource misallocation, and network-wide paralysis caused by local single-point failures. Therefore, there is an urgent need in this field for a novel method for large-scale cluster networking and dynamic maintenance to achieve objective intelligent networking and seamless topology evolution of highly dynamic node clusters with extremely low control overhead. Summary of the Invention

[0008] The purpose of this invention is to provide a large-scale UANET clustering networking method based on information entropy dynamic weighting and grace period model, so as to solve the technical problems faced in the existing technology of large-scale UAV cluster networking with more than 100 levels, such as extreme congestion of control signaling, difficulty in coordinating multi-dimensional heterogeneous states, and the ease with which single point failure can lead to network-wide paralysis.

[0009] To address the aforementioned technical problems, this invention provides a large-scale UANET clustering network method based on dynamic weighting of information entropy and a grace period model. The implementation steps strictly follow the system control logic and include the following steps: Step 1: Initialize the drone network and establish a cold start network based on static weights; Step 2: The supercluster head calculates the information entropy based on the state feature matrix and distributes the dynamic weight set; Step 3: Steady-state cluster reselection and primary / standby binding based on dynamic weights; Step 4: Multidimensional steady-state maintenance and seamless role switching based on grace period.

[0010] Furthermore, step 1 specifically includes the following steps: Step 11: Basic Neighbor Discovery. After all UAV nodes take off and go online, they are in an isolated state and have not yet joined any communication cluster. In this state, each UAV node periodically broadcasts neighbor discovery data packets through the omnidirectional antenna of its onboard communication module. The neighbor discovery data packets carry at least its own physical ID, GPS three-dimensional coordinates, remaining energy, and load status. Each UAV node receives data packets from other UAV nodes within a preset neighbor discovery period and records the sending nodes that meet the single-hop communication conditions as neighbor nodes, thereby constructing a neighbor state table at the link layer. Step 12: Load the default weights for cold start. At this point, the network has not yet established a supercluster head and is in the cold start phase. Each UAV node extracts the initial static weight set that is fixed in its local protocol stack by default. ; Step 13: Initial clustering and super cluster head establishment. Each UAV node participating in the initial clustering extracts data from its neighbor state table, substitutes it with the initial static weights to calculate its own steady-state comprehensive election weight score, and broadcasts the score to its single-hop neighbor nodes. The node with the highest score directly declares itself as the initial master cluster head. Subsequently, the initial master cluster heads of each region conduct a secondary interaction through a high-power cross-cluster control channel to elect the cluster head with the strongest initial comprehensive capability, which is then officially upgraded to the initial super cluster head. After the initial clustering is completed, the network forms a member layer, a cluster head layer, and a super cluster head layer. Among them, the member layer nodes are ordinary member UAV nodes that do not serve as master cluster heads or super cluster heads.

[0011] Furthermore, in step 12, during the initial network construction phase, in order to prioritize ensuring the rapid connectivity of the entire network, the initial static weight set sets the weight coefficient of the neighbor node degree to the maximum value; the neighbor node degree is the number of neighbor nodes that the UAV node detects within a preset neighbor discovery period and that meet the single-hop communication conditions.

[0012] Furthermore, the method for calculating the information entropy of the supercluster head in step 2 is as follows: The supercluster head sends a "status polling" signal to the entire network, collecting data on all drone nodes for the first time. The number of nodes is set to [number to be filled in]. The four-dimensional state data under actual flight conditions is constructed to a scale of State feature matrix : ; in, Represents the state feature matrix; This represents the total number of drone nodes currently participating in the statistics across the entire network; Indicates the first in the network The node at the th The true physical state value under the characteristic index, and , The four characteristic columns of the matrix correspond to the four-dimensional state parameters of the UAV node, respectively: remaining energy. Neighbor node degree Link quality and instantaneous load ; To eliminate the influence of different physical dimensions on election decisions, the extreme value normalization formula is used to normalize the matrix. Perform a dimensionless mapping to obtain the standard state matrix. Elements in the matrix satisfy: ; in, Indicates the first in the network The node at the th The standard value of the feature index after dimensionless mapping, and ; This indicates that within the current statistical period, the entire network... The node in the node The maximum value of the feature index; This indicates that within the current statistical period, the entire network... The node in the node The minimum value of a feature index; For matrix After performing extreme value normalization, the objective contribution ratio of each node's features is calculated. : Based on the principles of information theory, solve the... The output information entropy of the evaluation indicator : ; in, Indicates the first in the network The node at the th The proportion of a feature indicator in the total of all nodes for that feature indicator; Indicates the first The output information entropy of an evaluation indicator is used to measure the distribution of the difference of that indicator in the entire network data. The constant term is the term that represents the constant term. Calculate using the logarithm. This represents the total number of drone nodes participating in the statistics across the entire network, used to ensure the calculated information entropy. The value range is located at Between; when it appears When the boundary state is defined, it is specified ; Define the coefficient of difference for each indicator ; By normalizing the difference coefficient, the dynamic weight set for the current period is calculated: ; in, Indicates the first The coefficient of variation of the evaluation indicators is defined as follows: ; The larger the value, the more significant the difference in this indicator among all nodes in the network, and it should be given a higher evaluation weight. Indicates in At time 1, the supercluster head node uses the information entropy algorithm to... The final dynamic objective weight coefficients for the allocation of evaluation indicators, and satisfying the following conditions. ; The supercluster head will calculate the dynamic weight set. The dynamic weight set is encapsulated in global control signaling and broadcast to the UAV nodes participating in this round of cluster election. After receiving the dynamic weight set, each candidate UAV node replaces the initial static weight set with the dynamic weight set and calculates the steady-state comprehensive election weight score by combining it with local state parameters. This serves as the basis for electing the primary cluster leader and the backup cluster leader.

[0013] Furthermore, step 3 specifically includes the following steps: Step 31: Self-evaluation of election weights. After receiving the global policy control signaling from the supercluster head, the member-level drone nodes trigger the network to enter steady-state operation mode and perform a self-evaluation of election weights: the nodes extract the latest dynamic weight coefficients and combine them with the remaining energy maintained locally. Neighbor node degree Link quality and instantaneous load Substitute into the formula to calculate its own steady-state comprehensive election weight score. : ;in, Indicates the first Each member-level drone node The steady-state comprehensive election weight score calculated at each moment; to These correspond to the dynamic weighting coefficients of remaining energy, neighbor node degree, link quality, and load margin at the current moment, respectively. Indicates the first The current actual remaining power of each member-level drone node; Indicates the first The maximum physical battery capacity limit for each member-level drone node; Indicates the first The neighbor node degree of each member-level drone node; Indicates the first Instantaneous link quality assessment values ​​of each member-layer drone node after exponential moving average filtering and smoothing; Indicates the first The current length of the MAC layer transmit buffer queue of the airborne communication module of each member-level UAV node; Indicates the first The upper limit of the MAC layer send buffer queue capacity of the airborne communication module of each member-level UAV node; Step 32: Weight competition interlocking and main cluster head establishment, the opening length of each drone node is... During the election window, a node broadcasts its own information to its single-hop neighbor nodes. It also listens to the scores of neighboring nodes. After the time window expires, the node with the highest score within the single-hop neighbor range broadcasts the main cluster head declaration signal CH_CLAIM to establish its main cluster head status; the node that is not elected switches to a member-level drone node and sends the cluster entry request signal JOIN_REQ. Step 33: Establishing a backup cluster head and synchronizing asynchronous status. Within the same communication cycle in which the primary cluster head is established, the node ranked second in the cluster election score is designated as the backup cluster head. The backup cluster head enables promiscuous listening mode or receives periodic status packets from the primary cluster head, and synchronizes the cluster topology table and routing cache mirror in real time in the background to maintain a hot backup status.

[0014] Furthermore, the method for multidimensional steady-state maintenance in step 4 is as follows: When the master cluster head detects the occupancy rate of the MAC layer transmit buffer queue of the airborne communication module... Exceeding the high water level safety threshold At this time, the primary cluster head does not change the existing cluster topology and routing addressing relationship, but sends a load redirection activation command to the standby cluster head in hot standby state; the primary cluster head issues Redirect-REQ flow redirection signaling to some high-traffic source nodes, offloading subsequent service flows to the standby cluster head for cooperative forwarding; when the buffer queue falls back to the safe recovery threshold At that time, the single-point forwarding architecture is restored.

[0015] Furthermore, the method for seamless role switching based on the grace period in step 4 is as follows: When the remaining energy of the current master cluster leader is detected to have fallen below the critical threshold Or, the supercluster head may issue a new dynamic weight set, which could affect the current chief cluster head's overall election score. When a primary cluster head falls behind the current backup cluster head, the primary cluster head node automatically triggers a topology role handover procedure. During the handover initiation, this method strictly prohibits directly disconnecting existing physical links; instead, the primary cluster head activates a local queue with a duration of [duration missing]. A dynamic grace period timer is used; within this period, the current primary cluster head continues to maintain the underlying control channel connectivity to process existing data packets, but refuses to receive new service initialization requests; simultaneously, the primary cluster head serializes and encapsulates the latest local topology management sequence, dynamic routing mapping table, and time slot allocation table, and incrementally pushes them to the backup cluster head; this process continues until the state context transition is confirmed to be complete or... When the timer reaches zero, the standby cluster head immediately broadcasts a role reversal declaration to the entire network, officially becoming the new master cluster head and taking full control of routing and control, while the original master cluster head is smoothly downgraded to a regular member drone node.

[0016] By adopting the above technical solution, the present invention has the following beneficial effects: 1. Physical Congestion Prevention Architecture Based on Three-Layer Control Domain Isolation: This invention abandons the traditional flat or single-cluster network and constructs a three-layer physical hierarchy architecture consisting of a "member layer, cluster head layer, and super cluster head layer." Member nodes are only responsible for local environmental awareness and single-hop data reporting, and do not have cross-cluster routing control. Cluster head nodes are responsible for time slot allocation, information aggregation, and inter-cluster relay within their cluster. Super cluster head nodes are detached from the underlying high-frequency service flows and are dedicated to performing cross-regional macro-situational awareness and global policy distribution. Through this strict physical hierarchy isolation, this invention forcibly reduces the topology maintenance signaling, which originally explodes exponentially with the number of nodes in a large-scale network, and confines it to a local cluster. This completely cuts off the flood propagation path of invalid signaling at the lower level throughout the network, fundamentally eliminating the "broadcast storm" phenomenon across the entire network at the scale of hundreds or even thousands of nodes.

[0017] 2. Data-Driven Dynamic Weighting Mechanism Based on Information Entropy (EWM): Addressing the shortcomings of traditional clustering algorithms that rely on manual experience to solidify weights, this invention introduces an objective information entropy evaluation model. The supercluster head periodically aggregates the multi-dimensional heterogeneous states of all network nodes, including remaining energy, node degree, link quality, and load rate, constructing a state feature matrix. By calculating the distribution difference coefficient of each indicator among all network nodes (the larger the variance of an indicator across the entire network, the smaller its information entropy), the supercluster head node automatically calculates and assigns higher election weights to indicators with high differences. After receiving the dynamically distributed weights from the global database, UAV nodes substitute them into their local state parameters to complete the self-assessment and direct election of cluster head qualifications. This mechanism enables the network to possess extremely strong environmental adaptability; for example, when the cluster encounters strong winds causing uneven overall link quality, the weight of link quality will automatically surge. This dynamic weighting, entirely driven by current network data, accurately matches the core bottlenecks of a large-scale network at different flight stages, ensuring the optimal construction of the backbone topology.

[0018] 3. Primary / Backup Cooperative Load Offloading Mechanism under Non-Topology Change Conditions: Addressing the pain point of massive aggregated traffic easily causing instantaneous congestion at backbone nodes, this invention establishes a primary and backup dual-cluster head system during the networking phase and designs concurrent traffic distribution logic. In steady-state operation, the primary cluster head monitors the local buffer queue occupancy rate in real time. When the queue length exceeds the set high-water level safety threshold, but the node has not yet reached the energy depletion reconstruction condition, this method does not change the existing cluster topology, but directly activates the backup cluster head in hot backup state through control signaling. The primary cluster head uses flow redirection instructions to offload some low-priority service flows or data from specific source nodes to the backup cluster head for cooperative forwarding. This mechanism breaks through the physical capacity limit of a single cluster head, providing a flexible buffering method for large-scale concurrent communication. It achieves instantaneous and exponential expansion of network capacity without increasing topology maintenance overhead or changing existing routing relationships, effectively eliminating queuing delays and tail packet loss under sudden large traffic surges.

[0019] 4. Seamless handover mechanism for primary and backup cluster heads combined with the grace period model: To address topology oscillations caused by the failure of critical relay nodes, this invention proposes a "connect first, then disconnect" soft handover mechanism. When the remaining energy of the current primary cluster head falls below the critical safety threshold, or when its comprehensive score lags behind the backup cluster head due to global weight parameter updates, this method does not immediately disconnect the physical link, but instead activates a dynamic grace period. During this period, the primary cluster head continues to maintain connectivity of the underlying control channel to handle existing services, while smoothly migrating its cluster member list, time slot allocation table, and routing context to the backup cluster head after serialization. Only after state synchronization is completed or the grace period ends does the UAV node perform a formal role flip. This mechanism provides a crucial "oscillation damper" for the evolution of large-scale topologies, transforming the highly destructive "cliff-like disconnection" into a seamless "soft landing," effectively curbing the collective flooding and paging of hundreds of surrounding member-level nodes induced by the passive failure of a single cluster head, and completely avoiding the "avalanche-like" routing breaks and ineffective energy consumption that are prone to occur in large-scale cluster networks. Attached Figure Description

[0020] Figure 1 This is a flowchart of step 1 of the present invention; Figure 2 This is a flowchart of step 2 of the present invention; Figure 3 This is a flowchart of step 3 of the present invention; Figure 4 This is a flowchart of step 4 of the present invention. Detailed Implementation

[0021] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Those skilled in the art should understand that the following specific embodiments or implementation methods are a series of optimized configurations listed to further explain the specific content of the invention. These configuration methods can be combined or used in conjunction with each other, unless the invention explicitly states that some or a specific embodiment or implementation method cannot be associated with or used in conjunction with other embodiments or implementation methods. Furthermore, the following specific embodiments or implementation methods are merely optimized configurations and are not intended to limit the scope of protection of the invention.

[0023] The present invention will be further explained below with reference to specific embodiments. Example 1

[0024] This embodiment provides a large-scale UANET clustering network method based on dynamic weighting of information entropy and a grace period model. The implementation steps strictly follow the system control logic, and the specific execution flow is as follows: At a large-scale disaster site, 256 homogeneous multi-rotor UAVs (all UAV nodes equipped with the same communication modules, sensors, and battery capacity) were deployed to perform wide-area search missions. To ensure the stability and efficiency of the communication links, this method requires adaptively dividing these 256 UAVs into 16 communication clusters. The UAV swarm needs to maintain network connectivity and topology stability in complex weather conditions such as localized gusts and terrain conditions such as tall ruins obstructing the view.

[0025] like Figure 1 As shown, Step 1: Initialize the UAV network and cold start network construction based on static weights; Step 1 specifically includes the following steps: Step 11: Basic Neighbor Discovery: 256 homogeneous UAVs took off and dispersed towards the target area in a grid-like formation, in a state of waiting to be clustered before joining the communication cluster. Each node periodically broadcasts its current coordinates, initial battery level, etc., to its single-hop neighbor nodes via the omnidirectional antenna of its onboard communication module, and builds a neighbor state table at the link layer; Step 12: Load the default weights for cold start: At this point, the network has not yet established a supercluster head and is in a cold start phase; each drone node extracts the initial static weight set that is fixed in its local protocol stack by default. Step 13: Initial clustering and supercluster head establishment: Each UAV node participating in the initial clustering extracts data from its neighbor state table, substitutes it into the initial static weight to calculate its own steady-state comprehensive election weight score, and broadcasts the score to its single-hop neighbor nodes. The node with the highest score directly declares itself as the initial master cluster leader. Based on the location distribution and the initial number of connections, 16 optimal nodes whose physical locations are at the center of each region are quickly elected as the initial master cluster leaders, and the 256 UAVs are quickly divided into 16 communication clusters.

[0026] Subsequently, the initial master cluster heads of each region interacted a second time through a high-power cross-cluster control channel to select the cluster head with the strongest initial comprehensive capabilities, which was then officially upgraded to a super cluster head.

[0027] In step 12, during the initial network construction phase, in order to prioritize ensuring the rapid connectivity of the entire network, the initial static weight set sets the weight coefficient of the neighbor node degree to the maximum value; the neighbor node degree is the number of neighbor nodes that the UAV node detects within a preset neighbor discovery period and that meet the single-hop communication conditions.

[0028] like Figure 2 As shown, in step 2, the supercluster head calculates the information entropy based on the state feature matrix and distributes the dynamic weight set. As the gridded search progresses, environmental and task states lead to node heterogeneity. The primary node A of a certain cluster flies into a localized area of ​​strong winds. In order to maintain its hovering attitude, its motor output power surges, resulting in rapid power consumption. At the same time, node A is the first to discover a suspected target in the ruins, triggering the transmission of high-precision image data, which causes a sharp increase in the occupancy rate of its onboard communication module's MAC layer transmission buffer queue.

[0029] The remaining nodes in the cluster are either in a horizontal or targetless state. At this point, a dynamic weighting mechanism based on information entropy intervenes. The algorithm calculates local objective data and discovers that the dispersion of the "queue occupancy rate" and "remaining energy" indicators within the cluster has increased dramatically, indicating a decrease in information entropy and an increase in the effective information content of the indicators. Therefore, the algorithm automatically and dynamically increases the weights of these two indicators.

[0030] The method for calculating information entropy is as follows: The supercluster head sends a "status polling" signal to the entire network, collecting data on all drone nodes for the first time. The number of nodes is set to [number to be filled in]. The four-dimensional state data under actual flight conditions is constructed to a scale of State feature matrix : ; in, Represents the state feature matrix; This represents the total number of drone nodes currently participating in the statistics across the entire network; Indicates the first in the network The node at the th The true physical state value under the characteristic index, and , The four characteristic columns of the matrix correspond to the four-dimensional state parameters of the UAV node, respectively: remaining energy. Neighbor node degree Link quality and instantaneous load ; To eliminate the influence of different physical dimensions on election decisions, the extreme value normalization formula is used to normalize the matrix. Perform a dimensionless mapping to obtain the standard state matrix. Elements in the matrix satisfy: ; in, Indicates the first in the network The node at the th The standard value of the feature index after dimensionless mapping, and ; This indicates that within the current statistical period, the entire network... The node in the node The maximum value of the feature index; This indicates that within the current statistical period, the entire network... The node in the node The minimum value of a feature index; For matrix After performing extreme value normalization, the objective contribution ratio of each node's features is calculated. : Based on the principles of information theory, solve the... The output information entropy of the evaluation indicator : ; in, Indicates the first in the network The node at the th The proportion of a feature indicator in the total of all nodes for that feature indicator; Indicates the first The output information entropy of an evaluation indicator is used to measure the distribution of the difference of that indicator in the entire network data. The constant term is the term that represents the constant term. Calculate using the logarithm. This represents the total number of drone nodes participating in the statistics across the entire network, used to ensure the calculated information entropy. The value range is located at Between; when it appears When the boundary state is defined, it is specified ; Define the coefficient of difference for each indicator ; By normalizing the difference coefficient, the dynamic weight set for the current period is calculated: ; in, Indicates the first The coefficient of variation of the evaluation indicators is defined as follows: ; The larger the value, the more significant the difference in this indicator among all nodes in the network, and it should be given a higher evaluation weight. Indicates in At time 1, the supercluster head node uses the information entropy algorithm to... The final dynamic objective weight coefficients for the allocation of evaluation indicators, and satisfying the following conditions. ; The supercluster head will calculate the dynamic weight set. The dynamic weight set is encapsulated in global control signaling and broadcast to the UAV nodes participating in this round of cluster election. After receiving the dynamic weight set, each candidate UAV node replaces the initial static weight set with the dynamic weight set and calculates the steady-state comprehensive election weight score by combining it with local state parameters. This serves as the basis for electing the primary cluster leader and the backup cluster leader.

[0031] like Figure 3 As shown, step 3: steady-state cluster reselection and primary / backup binding based on dynamic weights; step 3 specifically includes the following steps: Step 31: Self-evaluation of election weights. After receiving the global policy control signaling from the supercluster head, the member-level drone nodes trigger the network to enter steady-state operation mode and perform a self-evaluation of election weights: the nodes extract the latest dynamic weight coefficients and combine them with the remaining energy maintained locally. Neighbor node degree Link quality and instantaneous load Substitute into the formula to calculate its own steady-state comprehensive election weight score. : ;in, Indicates the first Each member-level drone node The steady-state comprehensive election weight score calculated at each moment; to These correspond to the dynamic weighting coefficients of remaining energy, neighbor node degree, link quality, and load margin at the current moment, respectively. Indicates the first The current actual remaining power of each member-level drone node; Indicates the first The maximum physical battery capacity limit for each member-level drone node; Indicates the first The neighbor node degree of each member-level drone node; Indicates the first Instantaneous link quality assessment values ​​of each member-layer drone node after exponential moving average filtering and smoothing; Indicates the first The current length of the MAC layer transmit buffer queue of the airborne communication module of each member-level UAV node; Indicates the first The upper limit of the MAC layer send buffer queue capacity of the airborne communication module of each member-level UAV node; Step 32: Weight competition interlocking and main cluster head establishment, the opening length of each drone node is... During the election window, a node broadcasts its own information to its single-hop neighbor nodes. It also listens to the scores of neighboring nodes. After the time window expires, the node with the highest score within the single-hop neighbor range broadcasts the main cluster head declaration signal CH_CLAIM to establish its main cluster head status; the node that is not elected switches to a member-level drone node and sends the cluster entry request signal JOIN_REQ. Step 33: Establishing a backup cluster head and synchronizing asynchronous state. While node A is serving as the primary cluster head, node C within the same cluster (currently flying in a windless area with no large-volume data backhaul, idle queue, and sufficient power) is designated as the backup cluster head because its comprehensive score, calculated by the information entropy dynamic weighting model, ranks second within the cluster. The backup cluster head node C synchronizes the routing table of the primary cluster head A through periodic control signaling interaction, and synchronizes the cluster topology table and routing cache mirror in real time in the background to maintain a hot backup state.

[0032] like Figure 4 As shown, the method for multidimensional steady-state maintenance in step 4 is as follows: When the master cluster head node A detects the occupancy rate of the MAC layer transmit buffer queue of the airborne communication module... Exceeding the high water level safety threshold At this time, node A does not change the existing cluster topology and routing addressing relationship, but sends a load redirection activation command to the standby cluster head node C, which is in hot backup state; node A issues flow redirection signaling Redirect-REQ to some high-traffic source nodes, offloading subsequent service flows to node C for cooperative forwarding; when the buffer queue falls back to the safe recovery threshold At that time, the single-point forwarding architecture is restored.

[0033] Furthermore, the method for seamless role switching based on the grace period in step 4 is as follows: When the remaining energy of the current master cluster leader node A is detected to have fallen below the critical threshold Node A automatically triggers the topology role handover procedure; during the handover initiation, this method strictly prohibits directly disconnecting existing physical links, instead, Node A activates a local timeframe for a duration of [duration missing]. A dynamic grace period timer is set; within this grace period, node A continues to maintain control channel connectivity with member nodes within the cluster and the backup cluster head node C to process the received existing data packets and complete the incremental migration of topology management sequences, dynamic routing mapping tables, and time slot allocation tables; at the same time, node A rejects new service initialization requests.

[0034] Pending confirmation of state context transition completion or When the timer reaches zero, the standby cluster head node C broadcasts a role reversal declaration to all member nodes in the cluster, officially becoming the new primary cluster head and taking over routing and control. Node A smoothly degrades to a regular member node. Thus, the primary / standby cluster head switchover process is confined to the local cluster domain, avoiding topology oscillations caused by re-clustering.

[0035] The present invention has been described in detail above with reference to preferred embodiments and exemplary examples. However, it should be noted that these specific embodiments are merely illustrative explanations of the invention and do not constitute any limitation on the scope of protection of the invention. Various improvements, equivalent substitutions, or modifications can be made to the technical content and embodiments of the present invention without departing from the spirit and scope of protection of the invention, and all such modifications fall within the scope of protection of the present invention.

Claims

1. A large-scale UANET clustering networking method based on information entropy dynamic weighting and grace period model, characterized in that, The implementation steps are strictly carried out according to the system control logic, including the following steps: Step 1: Initialize the drone network and establish a cold start network based on static weights; Step 2: The supercluster head calculates the information entropy based on the state feature matrix and distributes the dynamic weight set; Step 3: Steady-state cluster reselection and primary / standby binding based on dynamic weights; Step 4: Multidimensional steady-state maintenance and seamless role switching based on grace period.

2. The large-scale UANET clustering network method based on dynamic weighting of information entropy and grace period model according to claim 1, characterized in that, Step 1 specifically includes the following steps: Step 11: Basic Neighbor Discovery. After all UAV nodes take off and go online, they are in an isolated state and have not yet joined any communication cluster. In this state, each UAV node periodically broadcasts neighbor discovery data packets through the omnidirectional antenna of its onboard communication module. The neighbor discovery data packets carry at least its own physical ID, GPS three-dimensional coordinates, remaining energy, and load status. Each UAV node receives data packets from other UAV nodes within a preset neighbor discovery period and records the sending nodes that meet the single-hop communication conditions as neighbor nodes, thereby constructing a neighbor state table at the link layer. Step 12: Load the default weights for cold start. At this point, the network has not yet established a supercluster head and is in the cold start phase. Each UAV node extracts the initial static weight set that is fixed in its local protocol stack by default. ; Step 13: Initial clustering and super cluster head establishment. Each UAV node participating in the initial clustering extracts data from its neighbor state table, substitutes it with the initial static weights to calculate its own steady-state comprehensive election weight score, and broadcasts the score to its single-hop neighbor nodes. The node with the highest score directly declares itself as the initial master cluster head. Subsequently, the initial master cluster heads of each region conduct a secondary interaction through a high-power cross-cluster control channel to elect the cluster head with the strongest initial comprehensive capability, which is then officially upgraded to the initial super cluster head. After the initial clustering is completed, the network forms a member layer, a cluster head layer, and a super cluster head layer. Among them, the member layer nodes are ordinary member UAV nodes that do not serve as master cluster heads or super cluster heads.

3. The large-scale UANET clustering network method based on dynamic weighting of information entropy and grace period model according to claim 2, characterized in that, In step 12, during the initial network construction phase, in order to prioritize ensuring the rapid connectivity of the entire network, the initial static weight set sets the weight coefficient of the neighbor node degree to the maximum value; the neighbor node degree is the number of neighbor nodes that the UAV node detects within a preset neighbor discovery period and that meet the single-hop communication conditions.

4. The large-scale UANET clustering network method based on dynamic weighting of information entropy and grace period model according to claim 1, characterized in that, The method for calculating the information entropy of the supercluster head in step 2 is as follows: The supercluster head sends a "status polling" signal to the entire network, collecting data on all drone nodes for the first time. The number of nodes is set to [number to be filled in]. The four-dimensional state data under actual flight conditions is constructed to a scale of State feature matrix : ; in, Represents the state feature matrix; This represents the total number of drone nodes currently participating in the statistics across the entire network; Indicates the first in the network The node at the th The true physical state value under the characteristic index, and , The four characteristic columns of the matrix correspond to the four-dimensional state parameters of the UAV node, respectively: remaining energy. Neighbor node degree Link quality and instantaneous load ; To eliminate the influence of different physical dimensions on election decisions, the extreme value normalization formula is used to normalize the matrix. Perform a dimensionless mapping to obtain the standard state matrix. Elements in the matrix satisfy: ; in, Indicates the first in the network The node at the th The standard value of the feature index after dimensionless mapping, and ; This indicates that within the current statistical period, the entire network... The node in the node The maximum value of the feature index; This indicates that within the current statistical period, the entire network... The node in the node The minimum value of a feature index; For matrix After performing extreme value normalization, the objective contribution ratio of each node's features is calculated. : Based on the principles of information theory, solve the... The output information entropy of the evaluation indicator : ; in, Indicates the first in the network The node at the th The proportion of a feature indicator in the total of all nodes for that feature indicator; Indicates the first The output information entropy of an evaluation indicator is used to measure the distribution of the difference of that indicator in the entire network data. The constant term is the term used to... Calculate using the logarithm. This represents the total number of drone nodes participating in the statistics across the entire network, used to ensure the calculated information entropy. The value range is located at Between; when it appears When the boundary state is defined, it is specified ; Define the coefficient of difference for each indicator ; By normalizing the difference coefficient, the dynamic weight set for the current period is calculated: ; in, Indicates the first The coefficient of variation of the evaluation indicators is defined as follows: ; The larger the value, the more significant the difference in this indicator among all nodes in the network, and it should be given a higher evaluation weight. Indicates in At time 1, the supercluster head node uses the information entropy algorithm to... The final dynamic objective weight coefficients for the allocation of evaluation indicators, and satisfying the following conditions. ; The supercluster head will calculate the dynamic weight set. The dynamic weight set is encapsulated in global control signaling and broadcast to the UAV nodes participating in this round of cluster election. After receiving the dynamic weight set, each candidate UAV node replaces the initial static weight set with the dynamic weight set and calculates the steady-state comprehensive election weight score by combining it with local state parameters. This serves as the basis for electing the primary cluster leader and the backup cluster leader.

5. The large-scale UANET clustering network method based on dynamic weighting of information entropy and grace period model according to claim 1, characterized in that, Step 3 specifically includes the following steps: Step 31: Self-evaluation of election weights. After receiving the global policy control signaling from the supercluster head, the member-level drone nodes trigger the network to enter steady-state operation mode and perform a self-evaluation of election weights: the nodes extract the latest dynamic weight coefficients and combine them with the remaining energy maintained locally. Neighbor node degree Link quality and instantaneous load Substitute into the formula to calculate its own steady-state comprehensive election weight score. : ;in, Indicates the first Each member-level drone node The steady-state comprehensive election weight score calculated at each moment; to These correspond to the dynamic weighting coefficients of remaining energy, neighbor node degree, link quality, and load margin at the current moment, respectively. Indicates the first The current actual remaining power of each member-level drone node; Indicates the first The maximum physical battery capacity limit for each member-level drone node; Indicates the first The neighbor node degree of each member-level drone node; Indicates the first Instantaneous link quality assessment values ​​of each member-layer drone node after exponential moving average filtering and smoothing; Indicates the first The current length of the MAC layer transmit buffer queue of the airborne communication module of each member-level UAV node; Indicates the first The upper limit of the MAC layer send buffer queue capacity of the airborne communication module of each member-level UAV node; Step 32: Weight competition interlocking and main cluster head establishment, the opening length of each drone node is... During the election window, a node broadcasts its own information to its single-hop neighbor nodes. It also listens to the scores of neighboring nodes. After the time window expires, the node with the highest score within the single-hop neighbor range broadcasts the main cluster head declaration signal CH_CLAIM to establish its main cluster head status; the node that is not elected switches to a member-level drone node and sends the cluster entry request signal JOIN_REQ. Step 33: Establishing a backup cluster head and synchronizing asynchronous status. Within the same communication cycle in which the primary cluster head is established, the node ranked second in the cluster election score is designated as the backup cluster head. The backup cluster head enables promiscuous listening mode or receives periodic status packets from the primary cluster head, and synchronizes the cluster topology table and routing cache mirror in real time in the background to maintain a hot backup status.

6. The large-scale UANET clustering network method based on dynamic weighting of information entropy and grace period model according to claim 1, characterized in that, The method for multidimensional steady-state maintenance in step 4 is as follows: When the master cluster head detects the occupancy rate of the MAC layer transmit buffer queue of the airborne communication module... Exceeding the high water level safety threshold At this time, the primary cluster head does not change the existing cluster topology and routing addressing relationship, but sends a load redirection activation command to the standby cluster head in hot standby state; the primary cluster head issues Redirect-REQ flow redirection signaling to some high-traffic source nodes, offloading subsequent service flows to the standby cluster head for cooperative forwarding; when the buffer queue falls back to the safe recovery threshold At that time, the single-point forwarding architecture is restored.

7. The large-scale UANET clustering network method based on dynamic weighting of information entropy and grace period model according to claim 1, characterized in that, The method for seamless role switching based on the grace period in step 4 is as follows: When the remaining energy of the current master cluster leader is detected to have fallen below the critical threshold Or, the supercluster head may issue a new dynamic weight set, which could affect the current chief cluster head's overall election score. When a primary cluster head falls behind the current backup cluster head, the primary cluster head node automatically triggers a topology role handover procedure. During the handover initiation, this method strictly prohibits directly disconnecting existing physical links; instead, the primary cluster head activates a local queue with a duration of [duration missing]. The dynamic grace period timer; within this period, the current master cluster head continues to maintain the underlying control channel connectivity to process existing data packets, but refuses to receive new service initialization requests; at the same time, the master cluster head serializes and encapsulates the latest local topology management sequence, dynamic routing mapping table and time slot allocation table, and incrementally pushes them to the standby cluster head; Pending confirmation of state context transition completion or When the timer reaches zero, the standby cluster head immediately broadcasts a role reversal declaration to the entire network, officially becoming the new master cluster head and taking full control of routing and control, while the original master cluster head is smoothly downgraded to a regular member drone node.

Images