A low-altitude intelligent networking adaptive clustering networking system for multi-dimensional resource cooperation

Abstract

The application provides a low-altitude intelligent networking adaptive clustering networking system facing multi-dimensional resource cooperation, and relates to the technical fields of low-altitude network communication and intelligent networking. The system comprises an information perception layer, which is used for acquiring surrounding node information, performing standardized processing and dynamic updating on various attribute parameters of the nodes, and obtaining an updated multi-dimensional neighbor table; an intelligent decision layer, which is used for calculating cluster head selection probability in a cluster head candidate set based on the updated multi-dimensional neighbor table, and obtaining a cluster head global update result; a cooperative operation layer, which is used for performing updating and calculation based on the cluster head global update result, and obtaining a transmission optimization result; and a fault-tolerant control layer, which is used for executing self-healing and topology recovery when a node fails or a network is interrupted based on the transmission optimization result, and obtaining a low-altitude intelligent networking adaptive clustering networking. The application solves the problems of low resource utilization efficiency, poor cluster structure stability, and lack of dynamic maintenance and optimization mechanism in the existing clustering networking scheme.

Description

Technical Field

[0001] This specification relates to the field of low-altitude network communication and intelligent networking technology, and in particular to a low-altitude intelligent network adaptive clustering networking system oriented towards multi-dimensional resource collaboration. Background Technology

[0002] With the rapid development of the low-altitude economy and the drone industry, the demand for communication, monitoring, navigation, and sensing services in low-altitude airspace is increasing, giving rise to the Low-Altitude Intelligent Network (LAIN), which is mainly composed of drones, airborne platforms, and ground nodes. LAIN can be widely used in various scenarios such as emergency communication, urban security, environmental monitoring, and air-ground collaborative sensing. However, the low-altitude space communication environment is complex, with dynamic changes in node altitude and channels susceptible to obstruction and interference, making it difficult to directly apply LAIN to traditional terrestrial communication network deployment schemes.

[0003] In LAIN, networking technology is crucial for ensuring communication reliability and system scalability. A well-designed networking mechanism enables efficient collaboration and data aggregation between nodes, reducing communication overhead and improving link stability, especially in situations with a large number of nodes, uneven energy distribution, and diverse mission types. Currently, UAV networking methods mainly include the following: First, a star-shaped networking method centered on a fixed ground base station. This method has a simple network structure and low control overhead, suitable for scenarios with a limited number of nodes or relatively fixed mission areas. However, in large-scale UAV deployments or when ground base station coverage is limited, communication bottlenecks and node access congestion problems easily occur, making it difficult to meet the high-dynamic, wide-area coverage requirements of low-altitude networking. Second, a planar self-organizing mode based on peer-to-peer connections. Nodes can spontaneously establish communication links based on indicators such as distance and signal strength, possessing strong flexibility and self-recovery capabilities. However, its mesh structure experiences significantly increased maintenance overhead and frequent route updates when the number of nodes is large. Third, a multi-level networking scheme based on clustering. Centralized management within clusters and collaborative communication between clusters are achieved through cluster head nodes, effectively reducing control signaling load and improving system scalability and energy efficiency.

[0004] However, current cluster-based networking solutions still face some problems and challenges:

[0005] (1) Cluster head selection is based on a single factor and lacks multi-dimensional resource integration: Existing low-altitude networking methods are mostly based on a single factor (such as node distance, remaining energy or signal strength) for cluster head selection and topology maintenance, failing to fully consider the coupling relationship of multi-dimensional resource elements such as energy, link quality, node load, and bandwidth, resulting in low overall network resource utilization efficiency;

[0006] (2) Lack of global optimization capability in algorithm performance: Some existing clustering algorithms have problems such as slow convergence speed, easy to get trapped in local optima and insufficient global search capability during the optimization process, making it difficult to achieve efficient resource scheduling and adaptive networking under complex time-varying topology conditions;

[0007] (3) Cluster structure is prone to instability, and networking and maintenance costs are high: In the low-altitude network environment with high-speed movement and frequent changes in tasks, the topology between nodes is constantly changing, and the cluster structure is prone to instability, resulting in a significant increase in networking and maintenance costs. Existing methods lack cluster head update and maintenance mechanisms, making it difficult to efficiently achieve continuous and stable management of the cluster structure. Summary of the Invention

[0008] To address the aforementioned shortcomings in existing technologies, this invention provides an adaptive clustering networking system for low-altitude intelligent networks oriented towards multi-dimensional resource collaboration. This system solves the problems of low resource utilization efficiency, poor cluster structure stability, and lack of dynamic maintenance and optimization mechanisms in existing clustering networking schemes.

[0009] To achieve the aforementioned objectives, the technical solution adopted by this invention is: a low-altitude intelligent network adaptive clustering networking system oriented towards multi-dimensional resource collaboration, comprising:

[0010] The information perception layer includes a neighbor discovery and state perception module and an attribute parameter evaluation and table maintenance module. The neighbor discovery and state perception module is used to obtain information about surrounding nodes and obtain a multi-dimensional neighbor node information table. The attribute parameter evaluation and table maintenance module is used to standardize and dynamically update the attribute parameters of each node based on the multi-dimensional neighbor node information table to obtain an updated multi-dimensional neighbor table.

[0011] The intelligent decision-making layer includes a cluster head candidate optimization calculation module and a cluster head election module. The cluster head candidate optimization calculation module is used to calculate the cluster head candidate set based on the updated multi-dimensional neighbor table using a multi-objective Pareto cluster head selection algorithm. The cluster head election module is used to calculate the cluster head selection probability in the cluster head candidate set, perform cluster head declaration and fast re-election, and obtain the global update result of the cluster head.

[0012] The collaborative operation layer includes a cluster member addition and table update module and a cross-cluster communication and autonomous system interaction module. The cluster member addition and table update module is used to perform cluster member access, utility calculation and cluster table update based on the global update result of the cluster head to obtain the cluster member update result; the cross-cluster communication and autonomous system interaction module is used to perform inter-cluster path calculation, data relay and resource interaction based on the cluster member update result to obtain the transmission optimization result.

[0013] The fault-tolerant control layer is used to perform self-healing and topology recovery when nodes fail or the network is interrupted, based on the transmission optimization results, to obtain the adaptive clustering of the low-altitude intelligent network.

[0014] The beneficial effects of the present invention are as follows: The present invention provides an adaptive clustering networking system for low-altitude intelligent networks oriented towards multi-dimensional resource collaboration. Through a combination of modularization and intelligent optimization methods, the resource adaptive clustering networking of low-altitude intelligent networks is realized. (1) The NSGA-II multi-objective Pareto optimization mechanism is used to comprehensively optimize candidate cluster head nodes in terms of energy consumption, link quality and load balancing, thereby improving energy efficiency and stability; (2) Node cognition and self-organization are realized through neighbor perception and multi-dimensional attribute tables; (3) A multi-layer fault tolerance mechanism is introduced to maintain stability even when there are node failures and frequent topology changes; (4) The system architecture has strong versatility and can be widely used in fields such as UAV swarm communication, low-altitude Internet of Things, and 6G air-ground collaborative networks.

[0015] Furthermore, the neighbor discovery and state awareness module includes:

[0016] The discovery unit is used by each node to periodically broadcast neighbor discovery messages to its communication radius, and each node receives the neighbor node message information.

[0017] The storage unit is used to receive and parse message information using neighboring nodes, and record the parsing results and received signal strength to the local storage unit;

[0018] The update unit is used to periodically compare the data in the storage unit with the records in the neighboring node information table using nodes. When new neighboring node information is detected, it is inserted into the neighboring node information table. For existing node records, their energy, coordinates, and timestamp fields are updated according to the latest received data to obtain the update result.

[0019] The link quality calculation unit is used to calculate the link quality based on the received signal strength indication, packet loss rate, and node distance.

[0020] ;

[0021] in, Indicates link quality. Represents a normalized mapping. Indicates the link Above, node The received signal strength indication is normalized during calculation. This indicates that the spatial distance between nodes is calculated using normalization. Indicates link packet loss rate , and Indicates the weighting factor. ,and ;

[0022] The generation unit is used to integrate the attribute information of all neighboring nodes based on the update results and the link quality index, and generate a multi-dimensional neighboring node information table.

[0023] This invention proposes a neighbor discovery and state awareness module and provides a link quality calculation method. By updating node information and calculating link quality in real time, it provides data support and optimization decision-making basis for the selection of cluster heads and the maintenance of network stability in low-altitude communication networks, effectively improving network reliability and communication quality.

[0024] Furthermore, the attribute parameter evaluation and table maintenance module includes:

[0025] The standardization unit is used to perform uniform standardization processing on multidimensional attribute parameters based on a multidimensional neighbor node information table to eliminate dimensional differences and obtain the standardization result.

[0026] The correction unit is used to dynamically correct the node state parameters based on the standardized results using a time-weighted update method, balancing the influence between historical values ​​and new observations to obtain the correction results.

[0027] The deletion unit is used to determine a node as a failed node when it fails to respond for multiple consecutive periods, and to delete the corresponding record from the neighboring node information table.

[0028] The maintenance unit is used to perform the longest prefix aggregation operation on nodes with the same task attributes or adjacent geographical locations based on the correction results, so as to reduce data redundancy and improve the efficiency of querying and maintaining node information, and obtain the updated multidimensional neighbor table.

[0029] This invention proposes an attribute parameter evaluation and table maintenance module. Through standardized processing and dynamic correction of multi-dimensional attribute parameters, it improves the accuracy and stability of node status in low-altitude communication networks. Furthermore, by promptly deleting failed nodes and performing longest prefix aggregation, it effectively reduces data redundancy and improves the efficiency of node information query and maintenance.

[0030] Furthermore, the cluster head candidate optimization calculation module includes:

[0031] The extraction unit is used to initialize the candidate node set and the number of candidate cluster heads based on the updated multidimensional neighbor table, and extract the multidimensional resource indicators of each node.

[0032] The optimization unit is used to perform parallel optimization of the three objective functions based on multidimensional resource indicators. It performs encoding on the initialized population, with each individual corresponding to a set of candidate cluster head node numbers. Tournament selection and crossover mutation operations are used to generate a new generation of population.

[0033] The generation unit is used to generate the next generation population based on the new generation population and according to the elite preservation strategy during the iteration process;

[0034] The election unit is used to select the group of nodes with the best overall performance from the current population, based on the cluster head election strategy, as the final cluster head candidate set.

[0035] This invention proposes a cluster head candidate optimization calculation module that considers the impact of multiple low-altitude resource dimensions on network stability. It optimizes cluster head election through a multi-objective optimization method, forming a cluster head candidate set, which provides a reliable candidate basis for subsequent cluster head elections and enhances the adaptability and stability of low-altitude communication networks.

[0036] Furthermore, the expression for the optimization of the three objective functions is as follows:

[0037] ;

[0038] in, Indicates cluster head energy consumption Indicates cluster head Its neighboring nodes Link quality between Indicates cluster head The load size, Represents a node movement speed, Indicates the speed threshold. This represents the total number of adjacent nodes to the cluster head. This represents the average energy consumption dimension of resources. This represents the total number of cluster heads in the network. This indicates the average link quality dimension of resources. This represents the average load dimension of resources. This represents the average load size of the cluster head, and j represents the index of the adjacent node to the cluster head;

[0039] The expression for the elite retention strategy is:

[0040] ;

[0041] in, This indicates that the elite player retains their actions. Indicates from the current population The best individual is selected by considering both individual rank and crowding degree (CD). Indicates population Offspring generated after crossover and mutation operations This represents the current generation population in the NSGA-II algorithm. For the next generation of the species.

[0042] This invention proposes a three-objective optimized cluster head selection expression, which comprehensively considers the impact of cluster head energy consumption, link quality, and load on cluster stability. At the same time, it introduces the cluster head movement speed as an optimization factor, which further improves the stability and adaptability of cluster head selection, thus providing theoretical support for ensuring network stability in complex dynamic low-altitude environments.

[0043] Furthermore, the cluster head election module includes:

[0044] The computational unit is used to calculate the probability of a cluster head becoming a cluster head based on the cluster head candidate set, using the fitness vector of the cluster head candidate node, its Pareto level, and its crowding distance.

[0045] The broadcast unit is used to broadcast cluster head announcement messages by each candidate cluster head node based on the selection probability. After receiving the announcement message, other ordinary nodes update their local cluster association table and establish logical connection relationships within the cluster.

[0046] The reselection unit is used to trigger a fast reselection mechanism when a normal node does not receive any cluster head announcement message within a preset time window. It re-performs probability sampling in the suboptimal solution set of the Pareto front generated in the previous round and selects a backup cluster head node from the suboptimal candidate set to ensure the continuous connectivity of the network and the stability of the cluster structure.

[0047] The replacement unit is used to periodically perform resource self-assessment using the selected cluster head node. When it detects that the remaining energy is lower than a set threshold or that the communication status is invalid, it automatically calls the local candidate cache to replace the cluster head and obtain the replacement result.

[0048] The global update unit is used to restart the multi-objective Pareto optimization algorithm in each running cycle based on the replacement result, so as to obtain the global update result of the cluster head.

[0049] This invention provides a cluster head election module that calculates the selection probability of a cluster head candidate node by utilizing the fitness vector, Pareto level, and congestion distance of the candidate node, and dynamically adjusts the cluster head node according to the network status, effectively enhancing the communication stability of the network and improving the reliability of the communication network in low-altitude areas.

[0050] Furthermore, the expression for the selection probability is:

[0051] ;

[0052] in, This represents the adjustable weighting coefficient. The first term calculated by the NSGA-II algorithm represents the result of... A crowded distance, This represents the resource expression for the average energy consumption dimension. This represents the resource expression for the average link quality dimension. This represents the resource expression for the average load dimension. Indicates the probability of selection. Represents node i in cluster c. Let j represent node j in cluster c, and K represent the total number of cluster heads in the network. The first term calculated by the NSGA-II algorithm represents the result of... A crowded distance.

[0053] A cluster head selection probability expression is provided, which jointly considers the energy consumption, link quality, load and speed factors of the cluster head, providing a theoretical basis for the sum-ordering of nodes in the cluster head election, improving the reliability of cluster head selection, and providing a reference for re-election.

[0054] Furthermore, the cluster member joining and table update module includes:

[0055] The sending unit is used to periodically broadcast JoinRequest messages using the cluster head node;

[0056] The evaluation unit is used to calculate the utility function based on the message information after the candidate node receives the cluster head broadcast, evaluate the comprehensive benefits of joining different cluster heads, and select the node with the highest utility as the assigned cluster head; wherein, the expression for the assigned cluster head is:

[0057]

[0058] in, Represents a node Add cluster head The utility value; Represents a node With cluster head Link quality metrics between them This indicates the maximum link quality between nodes; Indicates the spatial distance between the two. This represents the maximum distance between a node and the cluster head; and These represent the energy consumption of the cluster head and its load, respectively. and These represent the maximum node energy consumption and the maximum load, respectively. Indicates the adaptive weighting coefficient;

[0059] The determination unit is used to receive the joining application from the cluster head to which the candidate node belongs using the cluster head node, make an acceptance determination according to the preset capacity threshold, and record the acceptance result and member information in the cluster member information table;

[0060] The cluster member update unit is used to periodically broadcast cluster member table update messages based on the admission results and member information records, using the cluster head node to maintain data consistency within the cluster. When a member node is detected to have failed to respond to heartbeat detection or alarm messages for several consecutive periods, it is marked as a failed member and deleted from the member information table, thus obtaining the cluster member update result.

[0061] Furthermore, the cross-cluster communication and autonomous system interaction module includes:

[0062] The decision-making unit is used to periodically report status data to the autonomous system controller based on the cluster member update results and the autonomous system topology structure of each cluster head node. The autonomous system controller performs comprehensive evaluation and optimization based on the aggregated status data to generate the optimal resource scheduling decision.

[0063] The reconstruction unit is used to recalculate the cross-cluster path cost function in real time, triggering the cross-cluster relay reconstruction process to achieve seamless switching and continuous transmission of data streams, and obtain transmission optimization results; wherein, the expression of the cross-cluster path cost function is:

[0064] ;

[0065] in, Represents the cost function for cross-cluster paths. Representing a path The time delay, Representing a path The sum of the energy consumption of all nodes. Representing a path The reliability quantification index is positively correlated with the link quality along the path. These represent the weighting parameters for latency, energy consumption, and reliability, respectively.

[0066] Furthermore, the fault-tolerant control layer includes:

[0067] The fault unit is used to periodically detect the operating status of members within the cluster based on the transmission optimization results using the cluster head node. When no heartbeat information is received from a member within several consecutive detection cycles, or when the remaining energy of a node is detected to be lower than a set threshold, the cluster head node reports a fault event to the autonomous system controller.

[0068] The self-healing unit is used to execute a two-layer self-healing mechanism after detecting a fault event: intra-cluster rapid recovery: the cluster head node selects the member node with the highest comprehensive weight as the new cluster head according to the cached weight table to achieve local takeover and service continuation; cross-cluster recovery: the autonomous system controller recalculates the cross-cluster communication path and issues recovery instructions to the relevant cluster head nodes to maintain global network connectivity and service continuity, and obtains link stability and cluster head failure rate indicators.

[0069] The balancing unit is used to automatically perform cluster splitting or task migration operations when the network is under high load and the resource utilization of the cluster head node exceeds a preset threshold, so as to balance the load between clusters and reduce the interference level and obtain the current energy consumption level.

[0070] The dynamic adjustment unit is used to periodically collect statistics on the energy consumption level, link stability, and cluster head failure rate of the entire network. Based on the statistical results, it dynamically adjusts the weight parameters in the optimization algorithm to maintain the long-term stability and efficient operation of the network, thus obtaining the adaptive clustering networking of the low-altitude intelligent network. Attached Figure Description

[0071] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0072] Figure 1 This is an exemplary block diagram of a low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration, as shown in some embodiments of this specification.

[0073] Figure 2 This is an exemplary schematic diagram of a low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration, as shown in some embodiments of this specification. Detailed Implementation

[0074] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0075] Example

[0076] Figure 1 This is an exemplary block diagram of a low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration, as shown in some embodiments of this specification. Figure 1 and Figure 2 As shown, a low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration includes an information perception layer, an intelligent decision-making layer, a collaborative operation layer, and a fault-tolerant control layer connected in sequence.

[0077] The information perception layer includes a neighbor discovery and state perception module and an attribute parameter evaluation and table maintenance module. The neighbor discovery and state perception module is used to obtain information about surrounding nodes and obtain a multidimensional neighbor node information table. The attribute parameter evaluation and table maintenance module is used to standardize and dynamically update the attribute parameters of each node based on the multidimensional neighbor node information table to obtain an updated multidimensional neighbor table.

[0078] In some embodiments, the information perception layer includes a neighbor discovery and state perception module and an attribute parameter evaluation and table maintenance module. The neighbor discovery and state perception module is used to identify neighbors and collect state information between nodes. Through periodic broadcasting and message parsing mechanisms, it obtains basic attributes such as the unique identifier, location information, remaining energy, signal strength, and link quality of neighboring nodes, enabling dynamic perception of the local network environment. The attribute parameter evaluation and table maintenance module is used to normalize and time-weightedly update the collected multi-dimensional attribute parameters, completing the standardization of node states and maintaining the neighbor table. This ensures the timeliness and comparability of node information, providing reliable data support for subsequent cluster head candidate optimization and network decision-making.

[0079] In some embodiments, the neighbor discovery and state awareness module includes: a discovery unit, used to periodically broadcast neighbor discovery messages to the communication radius of each node, and each node receives neighbor node message information; a storage unit, used to receive and parse the message information using neighbor nodes, and record the parsing results and received signal strength in a local storage unit; an update unit, used to periodically compare the data in the storage unit with the records in the neighbor node information table using nodes, and insert new neighbor node information into the neighbor node information table when new neighbor node information is detected; for existing node records, the energy, coordinates and timestamp fields are updated according to the latest received data to obtain an update result; a link quality calculation unit, used to calculate the link quality based on the received signal strength indication, packet loss rate and node distance; and a generation unit, used to integrate the attribute information of all neighbor nodes based on the update result and the link quality index to generate a multi-dimensional neighbor node information table.

[0080] In some embodiments, the expression for the link quality index is:

[0081] ;

[0082] in, Indicates link quality. Represents a normalized mapping. Indicates the link Above, node The received signal strength indication is normalized during calculation. This indicates that the spatial distance between nodes is calculated using normalization. Indicates link packet loss rate , and Indicates the weighting factor. ,and .

[0083] In some embodiments, the neighbor discovery and state awareness module is used to identify neighbors, collect state data, and assess link quality among nodes. Each node periodically broadcasts neighbor discovery messages and establishes a neighbor information record table by receiving and parsing information sent by surrounding nodes. The system calculates the overall connection quality of neighboring nodes by comprehensively considering multi-dimensional parameters such as RSSI, LQI, spatial distance, and packet loss rate, thereby providing data support for the subsequent cluster head candidate evaluation and optimization calculation module.

[0084] The LQI value calculated by the above function reflects the overall reliability of links between nodes. The system uses this as an important basis for neighbor selection and topology maintenance, realizing quantitative perception of link status in the dynamic environment of low-altitude networks.

[0085] In some embodiments, the specific process of the neighbor discovery and state awareness module is as follows: S11: Nodes periodically broadcast "neighbor discovery messages" to the surrounding area. The message content includes fields such as node unique identifier (UID), location information, remaining energy level, and communication capability description. S12: Other nodes receive the message, parse and load it, and write the message information into the local storage unit to record the basic state of the current neighbor nodes. S13: Nodes compare the received new neighbor information with the local neighbor table. If it is a newly added node, a new record is inserted; if it is an existing node, its attribute fields and timestamp are updated to ensure the real-time validity of the neighbor table. S14: Based on RSSI and spatial distance, LQI is calculated, and a weighted function is used to comprehensively consider signal strength, packet loss rate, and distance attenuation to quantify the reliability of the link between neighbor nodes. S15: Finally, a neighbor table containing fields such as UID, energy, location information, link quality, and task attributes is formed, which is used for data support for subsequent cluster head candidate calculation and weight allocation. The main fields of the neighbor table are described in Table 1.

[0086] Table 1. Description of main fields in the neighbor table

[0087]

[0088] In some embodiments, the attribute parameter evaluation and table maintenance module includes: a standardization unit, used to perform unified standardization processing on multidimensional attribute parameters based on the multidimensional neighbor node information table to eliminate dimensional differences and obtain a standardization result; a correction unit, used to dynamically correct node state parameters based on the standardization result using a time-weighted update method to balance the influence between historical values ​​and new observations and obtain a correction result; a deletion unit, used to determine a node as a failed node when it has not responded for several consecutive periods and delete the corresponding record from the neighbor node information table; and a maintenance unit, used to perform longest prefix aggregation operation on nodes with the same task attributes or adjacent geographical locations based on the correction result to reduce data redundancy and improve the query and maintenance efficiency of node information and obtain an updated multidimensional neighbor table.

[0089] In some embodiments, the attribute parameter evaluation and table maintenance module is used to standardize, smooth, and dynamically maintain the attribute data of neighbor nodes to ensure unified comparison and stable updates of the system under multi-dimensional attributes. The module first normalizes parameters such as energy, link quality, signal strength, and packet loss rate to eliminate differences caused by different units, ensuring that multi-dimensional indicators can participate in unified weighted calculations. It then smooths the attribute parameters using a time-weighted method to reduce instantaneous fluctuations in node states and improve system stability and continuity. The module periodically checks the activity of nodes in the neighbor table. When a node fails to report information for several consecutive periods, it is marked as an invalid node and deleted to maintain the real-time accuracy of the neighbor table. For highly similar nodes located at the same access point or task layer, the system performs attribute aggregation or merged storage to reduce storage and maintenance overhead.

[0090] In some embodiments, the normalization formula is expressed as:

[0091] ;

[0092] in, This represents the normalization result. and This represents the minimum and maximum values ​​of attribute parameters (including energy, link quality, signal strength, and packet loss rate) obtained from the current neighbor table. This indicates the value of the node's own attribute parameters (including energy, link quality, signal strength, and packet loss rate).

[0093] In some embodiments, the expression for the attribute parameter is:

[0094] ;

[0095] in, As a smoothing factor, it controls the proportion of historical values ​​inherited. This represents the smoothed value from the previous period. This represents the instantaneous value of the new observation in this period. This is the normalized value of the attribute parameter calculated by the previous formula at the current moment.

[0096] In some embodiments, the specific process of the attribute parameter evaluation and table maintenance module is as follows: S21: Normalize the energy value, link quality, signal strength, and packet loss rate in the neighbor table according to the normalization formula to ensure that attributes with different dimensions can participate in unified weighting. S22: To avoid numerical jitter caused by frequent fluctuations, the normalized attribute values ​​are updated with time weighting smoothing according to the weighting formula. S23: If a neighbor node does not report any status information within T consecutive periods, the node is determined to be a failed node and is deleted from the neighbor table to maintain the timeliness and accuracy of neighbor information. S24: In the same access point or access layer, when a large number of nodes with highly similar attributes are detected, the system performs prefix aggregation at the storage end to merge and store nodes with similar task attributes or geographical locations into group entries, thereby reducing the computational and storage load of the autonomous system control layer.

[0097] The intelligent decision-making layer includes a cluster head candidate optimization calculation module and a cluster head election module. The cluster head candidate optimization calculation module is used to calculate the cluster head candidate set based on the updated multidimensional neighbor table using a multi-objective Pareto cluster head selection algorithm. The cluster head election module is used to calculate the selection probability of cluster heads in the cluster head candidate set, perform cluster head declaration and fast re-election, and obtain the global update result of cluster heads.

[0098] In some embodiments, the intelligent decision-making layer includes a cluster head candidate optimization calculation module and a cluster head election module. The cluster head candidate optimization calculation module constructs a multi-objective optimization model based on multi-dimensional resource attributes such as node energy, link quality, load level, and geographical distribution. It then uses the NSGA-II multi-objective Pareto optimization algorithm for global search and non-dominated ranking calculations to obtain the optimal set of cluster head candidates. The cluster head election module, based on the Pareto level, congestion distance, and adaptive probability distribution of candidate nodes, completes the final selection and broadcast announcement of the cluster head, enabling dynamic generation and stable maintenance of the cluster structure in the low-altitude network, and providing a decision-making basis for subsequent cluster member access and cross-cluster collaboration.

[0099] In some embodiments, the cluster head candidate optimization calculation module includes: an extraction unit, used to initialize the candidate node set and the number of candidate cluster heads based on the updated multidimensional neighbor table, and extract multidimensional resource indicators for each node; an optimization unit, used to perform parallel optimization of the three objective functions based on the multidimensional resource indicators, perform encoding on the initialized population, with each individual corresponding to a set of candidate cluster head node numbers, and generate a new generation population using tournament selection and crossover mutation operations; a generation unit, used to generate the next generation population based on the new generation population during the iteration process according to an elite retention strategy; and an election unit, used to select the set of nodes with the best overall performance from the current population according to a cluster head election strategy, as the final cluster head candidate set.

[0100] In some embodiments, the cluster head candidate optimization calculation module is used to achieve efficient multi-dimensional resource optimization calculation during the cluster head selection stage. Addressing the problem that traditional multi-factor weighted models are prone to getting trapped in local optima and have high result repetition rates, this invention introduces a dynamic search mechanism based on the NSGAII multi-objective Pareto optimization algorithm in the cluster head candidate stage. This mechanism combines non-dominated sorting with congestion distance evaluation to achieve a unification of global search and local optimization among multiple objectives such as energy consumption, link quality, and load balancing, thereby obtaining the optimality and stability of candidate cluster head selection. The system first extracts key attribute parameters of nodes based on the multi-dimensional neighbor information obtained by the "neighbor discovery and state awareness module," including remaining energy (E_total), link quality between adjacent nodes (LQI), mobility, and load imbalance. A multi-objective optimization model of the node is constructed using these parameters to measure the node's comprehensive capability as a cluster head. Its optimization objective is to minimize energy consumption, optimize link quality, and achieve load balancing while ensuring stable communication.

[0101] In some embodiments, the expression for the multi-objective optimization model is:

[0102] ;

[0103] in, Indicates cluster head energy consumption Indicates cluster head Its neighboring nodes Link quality between Indicates cluster head The load size, Represents a node movement speed, Indicates the speed threshold. This represents the total number of adjacent nodes to the cluster head. This represents the average energy consumption dimension of resources. This represents the total number of cluster heads in the network. This indicates the average link quality dimension of resources. This represents the average load dimension of resources. This represents the average load size of the cluster head, and j represents the index of the adjacent node to the cluster head.

[0104] In some embodiments, the NSGA-II algorithm is used to perform non-dominated sorting, crowding distance calculation, and elite retention operations on the objectives of the above multi-objective optimization model, thereby obtaining a Pareto optimal candidate cluster head set. The goal of the algorithm is to achieve the best trade-off between minimum energy consumption, optimal link quality, load balancing, and mobility stability.

[0105] In some embodiments, the expression for the elite retention strategy is:

[0106] ;

[0107] in, This indicates that the elite player retains their actions. Indicates from the current population The best individual is selected by considering both individual rank and crowding degree (CD). Indicates population Offspring generated after crossover and mutation operations This represents the current generation population in the NSGA-II algorithm. For the next generation of the species.

[0108] In some embodiments, the specific process of the cluster head candidate optimization calculation module is as follows: S31, Initialize the candidate node set. and the number of candidate cluster heads Extract multi-dimensional resource metrics for each node, including node energy consumption. Average link quality Node movement speed and intra-cluster load balancing coefficient The node's movement speed characterizes its spatial stability and is used as a constraint in the optimization process. A multi-objective optimization model is constructed to minimize the average energy consumption. Minimize load imbalance And maximize average link quality To optimize the objectives, the NSGA-II (Nondominated Sorting Genetic Algorithm II) algorithm is used to optimize the three objective functions in parallel. First, the initial population is encoded, with each individual corresponding to a set of candidate cluster head node numbers. Tournament selection and crossover / mutation operations are used to generate a new generation of the population. The population is then sorted according to the objective function values, and individuals are assigned values ​​based on their dominance level. And calculate the crowding distance of individuals. To maintain the diversity of the solution set, the sorting rule is: (a) Prioritize lower dominance levels. (b) When the levels are the same, priority should be given to choosing the distance from congestion. Larger individuals. S33. During the iteration process, the next generation of the population is generated according to the elite preservation strategy. The algorithm terminates when the number of iterations or the convergence threshold is met, obtaining the last generation of Pareto non-dominated solution set. .gather It contains several groups of cluster head candidate nodes with balanced performance, representing the optimal solution that compromises among multiple objective indicators; according to the cluster head election strategy, from... The group of nodes with the best overall performance is selected as the final cluster head candidate set. Priority determination can be based on a comprehensive scoring function. Calculate, when When a node reaches its maximum value or is located in the top 10% Pareto front region, the system marks it as an optimal cluster head candidate node and reports the result to the cluster head election module. Through this multi-objective Pareto optimization process, the system can achieve an optimal balance between energy consumption, link quality, and load balancing, thereby significantly improving the overall stability and energy efficiency of the network.

[0109] In some embodiments, the expression for the comprehensive scoring function is:

[0110] ;

[0111] in, and They represent the first The energy consumption and load size of each cluster head. Indicates the relationship with the first The link quality between each cluster head and its neighboring nodes. This represents the maximum link quality between the cluster head and its neighboring nodes. and These represent the maximum energy consumption and load size in the cluster head, respectively. Indicates the adaptive weighting coefficients, and .

[0112] In some embodiments, the cluster head election module includes: a calculation unit, used to calculate the probability of a cluster head becoming a cluster head based on the cluster head candidate set, using the fitness vector of the candidate node, its Pareto level, and its congestion distance; a broadcasting unit, used to broadcast cluster head announcement messages using each candidate cluster head node based on the selection probability, and other ordinary nodes update their local cluster association tables and establish intra-cluster logical connections after receiving the announcement messages; a re-election unit, used to trigger a fast re-election mechanism when an ordinary node does not receive any cluster head announcement messages within a preset time window, and re-perform probability sampling in the suboptimal solution set of the Pareto front generated in the previous round to select a backup cluster head node from the suboptimal candidate set to ensure the continuous connectivity of the network and the stability of the cluster structure; a replacement unit, used to periodically perform resource self-evaluation using the selected cluster head node, and automatically call the local candidate cache to replace the cluster head when it is detected that its remaining energy is lower than a set threshold or that the communication status is invalid, to obtain the replacement result; and a global update unit, used to restart the multi-objective Pareto optimization algorithm in each running cycle based on the replacement result to obtain the global update result of the cluster head.

[0113] In some embodiments, the expression for the selection probability is:

[0114] ;

[0115] in, This represents the adjustable weighting coefficient. The first term calculated by the NSGA-II algorithm represents the result of... A crowded distance, This represents the resource expression for the average energy consumption dimension. This represents the resource expression for the average link quality dimension. This represents the resource expression for the average load dimension. Indicates the probability of selection. Represents node i in cluster c. Let j represent node j in cluster c, and K represent the total number of cluster heads in the network. The first term calculated by the NSGA-II algorithm represents the result of... A crowded distance.

[0116] In some embodiments, the cluster head election module is used to determine the final set of cluster head nodes from the candidate cluster head set based on the Pareto optimization results and energy balance requirements of the nodes. This module outputs the optimal candidate set from the "cluster head candidate optimization calculation module". As input, each cluster head is dynamically determined through an adaptive probabilistic election mechanism based on Pareto level and crowding distance, ensuring the stability of the cluster structure, the balance of energy distribution, and the fairness of the election process. During the election process, to prevent similar nodes from occupying the cluster head position for extended periods, leading to uneven energy consumption, this embodiment introduces an adaptive probabilistic election model based on the node's Pareto level. distance from crowds A comprehensive calculation is performed to give multi-objective optimal nodes a higher probability of being elected, while retaining randomness to enhance the system's diversity and robustness. The system conducts a controlled random sampling in each election round, allowing nodes to dynamically rotate while maintaining energy balance, thereby achieving long-term load balancing and adaptive adjustment of cluster head energy.

[0117] In some embodiments, the cluster head election module specifically proceeds as follows: S41, based on the fitness vector of the cluster head candidate nodes... (Representing node energy consumption, link quality, and load balancing, respectively) and their Pareto level. distance from crowds Calculate the probability of it being selected as the cluster head. The system comprehensively considers node performance and diversity indicators, calculates probability allocation, and uses the resulting crowding distance to improve the uniformity of cluster head distribution; Individuals with lower probabilities automatically receive higher weights in probability calculations. S42: Each candidate cluster head node broadcasts a cluster head announcement message based on its selection probability. The message includes the node ID, remaining energy percentage, current load, cluster capacity limit, and timestamp information. Other ordinary nodes update their local cluster association tables after receiving the announcement message to establish logical connections within the cluster. S43: Ordinary nodes within a preset time window... If no cluster head announcement message is received, a fast reselection mechanism is triggered. Probabilistic sampling is then re-executed from the Pareto front suboptimal solution set generated in the previous round of NSGA-II. A backup cluster head node is selected from the (suboptimal candidate set) to ensure the continuous connectivity of the network and the stability of the cluster structure. S44. The selected cluster head node periodically performs a resource self-assessment; when it detects that its remaining energy is below a set threshold... If communication status fails, the local candidate cache will be automatically invoked. Perform cluster head replacement; S45, the system in each operating cycle The NSGA-II-based multi-objective Pareto optimization algorithm is restarted internally to achieve global cluster head updates, thereby maintaining the network's long-term optimization performance in dynamic environments.

[0118] The collaborative operation layer includes a cluster member joining and table update module and a cross-cluster communication and autonomous system interaction module. The cluster member joining and table update module is used to perform cluster member access, utility calculation and cluster table update based on the global update result of the cluster head to obtain the cluster member update result. The cross-cluster communication and autonomous system interaction module is used to perform inter-cluster path calculation, data relay and resource interaction based on the cluster member update result to obtain the transmission optimization result.

[0119] In some embodiments, the collaborative operation layer includes a cluster member joining and table updating module and a cross-cluster communication and autonomous system interaction module. The cluster member joining and table updating module enables ordinary nodes to access and register with cluster heads, calculates the optimal cluster head for a node using a utility function, and dynamically maintains and synchronizes the cluster member table to ensure the orderliness and data consistency of intra-cluster communication. The cross-cluster communication and autonomous system interaction module manages communication links and information coordination between different clusters, allocates and coordinates cross-cluster resources such as spectrum, time slots, and power according to autonomous system control policies, and is responsible for inter-cluster path maintenance, fault reconstruction, and domain-level data exchange, thereby ensuring global network connectivity and efficient collaborative operation.

[0120] In some embodiments, the cluster member joining and table update module includes: a sending unit, used to periodically broadcast JoinRequest messages using cluster head nodes; an evaluation unit, used to calculate a utility function based on message information after a candidate node receives the cluster head broadcast, evaluate the comprehensive benefits of joining different cluster heads, and select the node with the highest utility as the belonging cluster head; a determination unit, used to receive join requests from the belonging cluster heads of candidate nodes using cluster head nodes, make an acceptance determination based on a preset capacity threshold, and record the acceptance result and member information in the cluster member information table; and a cluster member update unit, used to periodically broadcast cluster member table update messages using cluster head nodes based on the acceptance result and member information records to maintain data consistency within the cluster; when a member node is detected to have failed to respond to heartbeat detection or alarm messages for multiple consecutive periods, it is marked as a failed member and deleted from the member information table to obtain the cluster member update result.

[0121] In some embodiments, the expression for the belonging cluster head is:

[0122]

[0123] in, Represents a node Add cluster head The utility value; Represents a node With cluster head Link quality metrics between them This indicates the maximum link quality between nodes; Indicates the spatial distance between the two. This represents the maximum distance between a node and the cluster head; and These represent the energy consumption of the cluster head and its load, respectively. and These represent the maximum node energy consumption and the maximum load, respectively. This represents the adaptive weighting coefficient.

[0124] In some embodiments, the cluster member joining and table update module is used to enable dynamic access and cluster member information maintenance for ordinary nodes after cluster head election. To ensure the self-organization and stability of the cluster structure, the system designs a membership mechanism based on a utility function: after the cluster head node completes its election, it actively broadcasts a "join invitation message" to its neighboring nodes. The message contains information such as the cluster head identifier, energy status, cluster load threshold, and access restrictions. After receiving invitations from multiple cluster heads, ordinary nodes calculate the utility value based on indicators such as link quality, cluster load, and communication cost, and select the optimal cluster head to join, thereby achieving autonomous membership and balanced distribution within the cluster. Nodes select members based on the utility function calculation results. The largest cluster head is designated as the host node, and a join confirmation message is sent to it. This method ensures node autonomy in joining while balancing intra-cluster communication reliability and energy balance, avoiding the communication burden and link congestion caused by centralized cluster formation. When a cluster head receives a join request, it decides whether to accept the request based on the current capacity threshold and remaining energy, and records the member's identifier (ID), task attributes, and last update time in its local cluster member table. The system maintains the consistency of cluster member information through a periodic broadcast cluster table update mechanism, thereby maintaining the long-term stability and dynamic self-maintenance capability of the cluster structure.

[0125] In some embodiments, the specific process of the cluster member joining and table update module is as follows: S51: After the cluster head node election is completed, the system starts the distributed member joining process. The cluster head node broadcasts a "JoinRequest" message to its surrounding neighbors, which includes information such as cluster head identifier, energy status, cluster load limit, and access conditions. S52: After receiving one or more cluster head invitations, ordinary nodes calculate the utility function based on the attributes of each cluster head. And select the cluster head with the highest utility value. As the host node, a join confirmation request is sent to it. S53: Upon receiving a member join request, if the cluster capacity has not reached its limit and energy conditions permit, the cluster head node accepts the member node and writes its information into the cluster member table. The member table includes member ID, task attributes, communication status, and update timestamp; when a member is offline or has no heartbeat response for a long time, the system automatically deletes the corresponding record from the table to maintain the real-time and accuracy of cluster member information. S54: The cluster head node periodically broadcasts cluster member table update packets. After receiving them, nodes within the cluster update their own caches to ensure that each member node maintains a consistent cluster structure view. When the number of cluster members or the task load changes significantly, the system will trigger a cluster reconfiguration mechanism to maintain communication balance and topology stability.

[0126] In some embodiments, the cross-cluster communication and autonomous system interaction module includes: a decision-making unit, used to periodically report status data to the autonomous system controller based on the cluster member update results and the autonomous system topology of each cluster head node; the autonomous system controller performs comprehensive evaluation and optimization based on the aggregated status data to generate the optimal resource scheduling decision; and a reconstruction unit, used to recalculate the cross-cluster path cost function in real time, trigger the cross-cluster relay reconstruction process, realize seamless switching and continuous transmission of data streams, and obtain transmission optimization results.

[0127] In some embodiments, the expression for the cross-cluster path cost function is:

[0128] ;

[0129] in, Represents the cost function for cross-cluster paths. Representing a path The time delay, Representing a path The sum of the energy consumption of all nodes. Representing a path The reliability quantification index is positively correlated with the link quality along the path. These represent the weighting parameters for latency, energy consumption, and reliability, respectively.

[0130] In some embodiments, the cross-cluster communication and autonomous system interaction module is used to manage inter-cluster communication, resource coordination, and domain-level data exchange. When different clusters need to transmit data or share control information, the system dynamically establishes cross-cluster links based on network topology and resource distribution. To improve the efficiency of cross-cluster communication, this invention introduces a joint optimization strategy of shortest path and energy constraints. By calculating the cost function of candidate paths, the system selects the cross-cluster relay path with the best overall performance, achieving low-latency and high-reliability data interaction. The system traverses all candidate cross-cluster paths, calculates their cost value, and selects the path with the minimum cost as the optimal cross-cluster transmission path, achieving a dual minimization of latency and energy consumption.

[0131] In some embodiments, the specific process of the cross-cluster communication and autonomous system interaction module is as follows:

[0132] S61: When a communication requirement between clusters is detected, the system first determines whether a stable relay link has been established between adjacent clusters; if not, it collects link information from adjacent clusters through the neighborhood control module and generates a candidate path set. S62: The system then uses the cost function... Each path is calculated and sorted, and the path with the lowest cost is selected as the final cross-cluster communication channel for data transmission and control command distribution. S63: At the autonomous system interaction level, each cluster head node periodically reports key parameters such as energy consumption status, link quality, and fault alarms to the autonomous system controller. The controller adjusts domain-level resource allocation and routing strategies accordingly to achieve global resource coordination and redundancy protection. S64: When a cluster head or link fails, the system immediately triggers a rapid reconfiguration mechanism, automatically selecting a new cross-cluster path to ensure uninterrupted communication, thereby achieving fault tolerance and self-recovery within the autonomous system.

[0133] The fault-tolerant control layer is used to perform self-healing and topology recovery when nodes fail or the network is interrupted, based on the transmission optimization results, to obtain the adaptive clustering of the low-altitude intelligent network.

[0134] In some embodiments, the fault-tolerant control layer is used to cope with abnormal situations such as cluster head failure, link breakage, and energy depletion, ensuring the long-term stable operation and data transmission continuity of the low-altitude intelligent network. When a cluster head node cannot continue to maintain intra-cluster communication due to energy decay, hardware damage, or channel interruption, the system can trigger a rapid recovery and self-healing mechanism locally or globally. By periodically detecting the intra-cluster internal beat signal and energy threshold, the node's survival status is determined, and self-organized reconstruction is achieved according to a preset two-layer recovery strategy—local-recovery and inter-cluster recovery. The former is used to quickly replace the failed cluster head to maintain stable intra-cluster communication, while the latter is used to recalculate the inter-cluster path and restore autonomous system-level data connectivity when multiple cluster links break. In the inter-cluster reconstruction phase, the system uses a multi-objective path cost function to select the recovery path to minimize latency and energy consumption and maximize link reliability.

[0135] In some embodiments, the fault-tolerant control layer includes: a fault unit, configured to periodically detect the operating status of cluster members based on transmission optimization results using cluster head nodes; when no heartbeat information is received from a member within multiple consecutive detection periods, or when the remaining energy of a node is detected to be lower than a set threshold, the cluster head node reports a fault event to the autonomous system controller; and a self-healing unit, configured to execute a two-layer self-healing mechanism after detecting a fault event: intra-cluster fast recovery: the cluster head node selects the member node with the highest comprehensive weight as the new cluster head according to the cached weight table to achieve partial takeover and service continuation; cross-cluster recovery: the autonomous system controller recalculates the cross-cluster communication path and... The system sends recovery commands to relevant cluster head nodes to maintain global network connectivity and service continuity, obtaining link stability and cluster head failure rate indicators. A balancing unit automatically performs cluster splitting or task migration operations when the network is under high load and the resource utilization of cluster head nodes exceeds a preset threshold, balancing inter-cluster load and reducing interference levels, thus obtaining the current energy consumption level. A dynamic adjustment unit periodically calculates the energy consumption level, link stability, and cluster head failure rate indicators of the entire network, and dynamically adjusts the weight parameters in the optimization algorithm based on the statistical results to maintain long-term network stability and efficient operation, resulting in an adaptive clustering network for the Low-Altitude Intelligent Network.

[0136] In some embodiments, the specific process of the system recovery and fault tolerance module is as follows: S71: The system periodically receives heartbeat packets reported by cluster members. When a cluster head node fails to receive a heartbeat packet from a specified member for several consecutive periods, it determines that the node has failed and removes it from the member table; if the cluster head's energy is below a threshold or the announcement packet is abnormally lost, the local fast election mechanism of neighboring nodes is triggered to complete the replacement, and a recovery event is reported to the upper-layer control module at the same time. S72: The system executes a two-layer recovery strategy according to the fault type: ① Local-Recovery: When the cluster head failure causes the intra-cluster communication to be interrupted, the system quickly determines a new cluster head based on the previous cluster head optimization results, realizing intra-cluster reconstruction and task inheritance. This process is a local autonomous operation and can complete topology repair within seconds; ② Inter-Cluster Recovery: When multiple cluster heads fail or inter-cluster links break, the autonomous system control layer recalculates the inter-cluster path cost function. The optimal path is selected to re-establish inter-cluster links, achieving domain-level data recovery and transmission reconnection. S73: When cluster head load or energy utilization exceeds the upper limit threshold, When the cluster energy level is below a set threshold, the Autonomous System Controller (AS / RS) automatically triggers a cluster splitting mechanism, prioritizing the establishment of new clusters using neighboring candidate nodes and migrating some members to achieve load sharing and energy balancing. If the cluster energy level remains below a set threshold for an extended period, a task transfer mechanism is triggered, automatically migrating critical tasks to nearby high-energy nodes to maintain continuous network operation. S74: The AS / RS periodically summarizes the operational status parameters of each cluster, including the number of members, average energy, link stability, and failure rate, and dynamically adjusts the weight parameters of the optimization algorithm based on the monitoring results. When uneven energy distribution or increased link fluctuations are detected within a cluster, the control layer will trigger resource reallocation and parameter correction operations, such as: increasing the load protection weight of low-energy nodes; dynamically adjusting communication power to balance interference; and issuing cluster structure reconstruction commands to restore topology consistency.

[0137] Through the aforementioned mechanisms, the system can achieve multi-level self-healing and fault-tolerant scheduling, maintaining network connectivity and stability in the event of local anomalies or energy imbalances. This scheme effectively prevents link interruptions and concentrated energy consumption caused by single-point failures of cluster heads, thereby significantly improving the robustness and service continuity of the low-altitude intelligent network in complex dynamic environments.

[0138] This method can be widely applied to scenarios such as low-altitude UAV communication, low-altitude monitoring, emergency rescue, and air-ground collaborative sensing. It can be used to achieve efficient and stable UAV adaptive networking and energy optimization management in low-altitude environments with dynamic changes in multiple nodes, complex channel conditions, and limited energy, thereby improving the network's resource utilization, topology robustness, and communication reliability.

[0139] In some embodiments of this specification, an adaptive clustering networking system for low-altitude intelligent networks oriented towards multi-dimensional resource collaboration is provided. Through a combination of modularization and intelligent optimization methods, resource adaptive clustering networking of low-altitude intelligent networks is realized. (1) The NSGA-II multi-objective Pareto optimization mechanism is used to comprehensively optimize candidate cluster head nodes in terms of energy consumption, link quality and load balancing, thereby improving energy efficiency and stability; (2) Node cognition and self-organization are realized through neighbor perception and multi-dimensional attribute tables; (3) A multi-layer fault tolerance mechanism is introduced to maintain stability even when there are node failures and frequent topology changes; (4) The system architecture has strong versatility and can be widely used in fields such as UAV swarm communication, low-altitude Internet of Things, and 6G air-ground collaborative networks.

Claims

1. A low-altitude intelligent networking adaptive clustering networking system oriented to multi-dimensional resource collaboration, characterized in that, include: The information perception layer includes a neighbor discovery and state perception module and an attribute parameter evaluation and table maintenance module. The neighbor discovery and state perception module is used to obtain information about surrounding nodes and obtain a multi-dimensional neighbor node information table. The attribute parameter evaluation and table maintenance module is used to standardize and dynamically update the attribute parameters of nodes based on the multidimensional neighbor node information table, so as to obtain the updated multidimensional neighbor table. The intelligent decision-making layer includes a cluster head candidate optimization calculation module and a cluster head election module. The cluster head candidate optimization calculation module is used to calculate the cluster head candidate set based on the updated multi-dimensional neighbor table using a multi-objective Pareto cluster head selection algorithm. The cluster head election module is used to calculate the cluster head selection probability in the cluster head candidate set, perform cluster head declaration and fast re-election, and obtain the global update result of the cluster head. The cluster head candidate optimization calculation module includes: The extraction unit is used to initialize the candidate node set and the number of candidate cluster heads based on the updated multidimensional neighbor table, and extract the multidimensional resource indicators of each node. The optimization unit is used to perform parallel optimization of the three objective functions based on multidimensional resource indicators. It performs encoding on the initialized population, with each individual corresponding to a set of candidate cluster head node numbers. Tournament selection and crossover mutation operations are used to generate a new generation of population. The expression for optimizing the three objective functions is as follows: ； in, Indicates cluster head energy consumption Indicates cluster head Its neighboring nodes Link quality between Indicates cluster head The load size, Represents a node movement speed, Indicates the speed threshold. This represents the total number of adjacent nodes to the cluster head. This represents the average energy consumption dimension of resources. This represents the total number of cluster heads in the network. This indicates the average link quality dimension of resources. This represents the average load dimension of resources. This represents the average load size of the cluster head, and j represents the index of the adjacent node to the cluster head; The generation unit is used to generate the next generation population based on the new generation population and according to the elite preservation strategy during the iteration process; The election unit is used to select the group of nodes with the best overall performance from the current population according to the cluster head election strategy, as the final cluster head candidate set; The collaborative operation layer includes a cluster member addition and table update module and a cross-cluster communication and autonomous system interaction module. The cluster member addition and table update module is used to perform cluster member access, utility calculation and cluster table update based on the global update result of the cluster head to obtain the cluster member update result; the cross-cluster communication and autonomous system interaction module is used to perform inter-cluster path calculation, data relay and resource interaction based on the cluster member update result to obtain the transmission optimization result. The fault-tolerant control layer is used to perform self-healing and topology recovery when nodes fail or the network is interrupted, based on the transmission optimization results, to obtain the adaptive clustering of the low-altitude intelligent network.

2. The low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration as described in claim 1, characterized in that, The neighbor discovery and status awareness module includes: The discovery unit is used by each node to periodically broadcast neighbor discovery messages to its communication radius, and each node receives the neighbor node message information. The storage unit is used to receive and parse message information using neighboring nodes, and record the parsing results and received signal strength to the local storage unit; The update unit is used to periodically compare the data in the storage unit with the records in the neighboring node information table using nodes. When new neighboring node information is detected, it is inserted into the neighboring node information table. For existing node records, their energy, coordinates, and timestamp fields are updated according to the latest received data to obtain the update result. The link quality calculation unit is used to calculate the link quality based on the received signal strength indication, packet loss rate, and node distance. ； in, Indicates link quality. Represents a normalized mapping. Indicates the link Above, node The received signal strength indication is normalized during calculation. This indicates that the spatial distance between nodes is calculated using normalization. Indicates link packet loss rate, , and Indicates the weighting factor. ,and ; The generation unit is used to integrate the attribute information of all neighboring nodes based on the update results and the link quality index, and generate a multi-dimensional neighboring node information table.

3. The low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration as described in claim 1, characterized in that, The attribute parameter evaluation and table maintenance module includes: The standardization unit is used to perform uniform standardization processing on multidimensional attribute parameters based on a multidimensional neighbor node information table to eliminate dimensional differences and obtain the standardization result. The correction unit is used to dynamically correct the node state parameters based on the standardized results using a time-weighted update method, balancing the influence between historical values ​​and new observations to obtain the correction results. The deletion unit is used to determine a node as a failed node when it fails to respond for multiple consecutive periods, and to delete the corresponding record from the neighboring node information table. The maintenance unit is used to perform the longest prefix aggregation operation on nodes with the same task attributes or adjacent geographical locations based on the correction results, so as to reduce data redundancy and improve the efficiency of querying and maintaining node information, and obtain the updated multidimensional neighbor table.

4. The low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration according to claim 1, characterized in that, The expression for the elite retention strategy is: ； in, This indicates that the elite player retains their actions. Indicates from the current population The best individual is selected by considering both individual ranking and crowding. Indicates population Offspring generated after crossover and mutation operations This represents the current generation population in the NSGA-II algorithm. For the next generation of the species.

5. The low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration according to claim 1, characterized in that, The cluster head election module includes: The computational unit is used to calculate the probability of a cluster head becoming a cluster head based on the cluster head candidate set, using the fitness vector of the cluster head candidate node, its Pareto level, and its crowding distance. The broadcast unit is used to broadcast cluster head announcement messages by each candidate cluster head node based on the selection probability. After receiving the announcement message, other ordinary nodes update their local cluster association table and establish logical connection relationships within the cluster. The reselection unit is used to trigger a fast reselection mechanism when a normal node does not receive any cluster head announcement message within a preset time window. It re-performs probability sampling in the suboptimal solution set of the Pareto front generated in the previous round and selects a backup cluster head node from the suboptimal candidate set to ensure the continuous connectivity of the network and the stability of the cluster structure. The replacement unit is used to periodically perform resource self-assessment using the selected cluster head node. When it detects that the remaining energy is lower than a set threshold or that the communication status is invalid, it automatically calls the local candidate cache to replace the cluster head and obtain the replacement result. The global update unit is used to restart the multi-objective Pareto optimization algorithm in each running cycle based on the replacement result, so as to obtain the global update result of the cluster head.

6. The low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration as described in claim 5, characterized in that, The expression for the selection probability is: ； in, This represents the adjustable weighting coefficient. The first term calculated by the NSGA-II algorithm represents the result of... A crowded distance, This represents the resource expression for the average energy consumption dimension. This represents the resource expression for the average link quality dimension. This represents the resource expression for the average load dimension. Indicates the probability of selection. Represents node i in cluster c. Let j represent node j in cluster c, and K represent the total number of cluster heads in the network. The first term calculated by the NSGA-II algorithm represents the result of... A crowded distance.

7. The low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration according to claim 1, characterized in that, The cluster member addition and table update module includes: The sending unit is used to periodically broadcast JoinRequest messages using the cluster head node; The evaluation unit is used to calculate the utility function based on the message information after the candidate node receives the cluster head broadcast, evaluate the comprehensive benefits of joining different cluster heads, and select the node with the highest utility as the assigned cluster head; wherein, the expression for the assigned cluster head is: in, Represents a node Add cluster head The utility value; Represents a node With cluster head Link quality metrics between them This indicates the maximum link quality between nodes; Indicates the spatial distance between the two. This represents the maximum distance between a node and the cluster head; and These represent the energy consumption of the cluster head and its load, respectively. and These represent the maximum node energy consumption and the maximum load, respectively. Indicates the adaptive weighting coefficient; The determination unit is used to receive the joining application from the cluster head to which the candidate node belongs using the cluster head node, make an acceptance determination according to the preset capacity threshold, and record the acceptance result and member information in the cluster member information table; The cluster member update unit is used to periodically broadcast cluster member table update messages based on the admission results and member information records, using the cluster head node to maintain data consistency within the cluster. When a member node is detected to have failed to respond to heartbeat detection or alarm messages for several consecutive periods, it is marked as a failed member and deleted from the member information table, thus obtaining the cluster member update result.

8. The low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration according to claim 1, characterized in that, The cross-cluster communication and autonomous system interaction module includes: The decision-making unit is used to periodically report status data to the autonomous system controller based on the cluster member update results and the autonomous system topology structure of each cluster head node. The autonomous system controller performs comprehensive evaluation and optimization based on the aggregated status data to generate the optimal resource scheduling decision. The reconstruction unit is used to recalculate the cross-cluster path cost function in real time, triggering the cross-cluster relay reconstruction process to achieve seamless switching and continuous transmission of data streams, and obtain transmission optimization results; wherein, the expression of the cross-cluster path cost function is: ； in, Represents the cross-cluster path cost function. Representing a path The time delay, Representing a path The sum of the energy consumption of all nodes. Representing a path The reliability quantification index is positively correlated with the link quality along the path. These represent the weighting parameters for latency, energy consumption, and reliability, respectively.

9. The low-altitude intelligent network adaptive clustering networking system for multi-dimensional resource collaboration according to claim 1, characterized in that, The fault-tolerant control layer includes: The fault unit is used to periodically detect the operating status of members within the cluster based on the transmission optimization results using the cluster head node. When no heartbeat information is received from a member within several consecutive detection cycles, or when the remaining energy of a node is detected to be lower than a set threshold, the cluster head node reports a fault event to the autonomous system controller. The self-healing unit is used to execute a two-layer self-healing mechanism after detecting a fault event: intra-cluster rapid recovery: the cluster head node selects the member node with the highest comprehensive weight as the new cluster head according to the cached weight table to achieve local takeover and service continuation; cross-cluster recovery: the autonomous system controller recalculates the cross-cluster communication path and issues recovery instructions to the relevant cluster head nodes to maintain global network connectivity and service continuity, and obtains link stability and cluster head failure rate indicators. The balancing unit is used to automatically perform cluster splitting or task migration operations when the network is under high load and the resource utilization of the cluster head node exceeds a preset threshold, so as to balance the load between clusters and reduce the interference level and obtain the current energy consumption level. The dynamic adjustment unit is used to periodically collect statistics on the energy consumption level, link stability, and cluster head failure rate of the entire network. Based on the statistical results, it dynamically adjusts the weight parameters in the optimization algorithm to maintain the long-term stability and efficient operation of the network, thus obtaining the adaptive clustering networking of the low-altitude intelligent network.

Images