Cluster-based unmanned swarm ad hoc network system

Abstract

The application relates to the technical field of wireless communication networks, and particularly discloses a cluster-based unmanned cluster ad hoc network system, which comprises a distributed spectrum situation awareness module, a dynamic clustering and cluster head election module, a spectrum-topology joint optimization routing module and a cross-layer cooperative control module. Through cooperative sensing of an electromagnetic environment, real-time spectrum quality is integrated into cluster head election and routing optimization, and closed-loop linkage and adaptive reconstruction of the three are realized, so that the spectrum utilization, link reliability and overall communication performance of the network in a dynamic environment are improved. The system introduces the distributed spectrum situation awareness module, realizes high-resolution and low-latency cooperative sensing of a complex dynamic electromagnetic environment, and provides real-time and accurate spectrum situation information for network decision-making.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication network technology, specifically relating to a cluster-based unmanned cluster self-organizing network system. Background Technology

[0002] Unmanned cluster ad hoc networking technology, as an important branch of modern communication networks, aims to enable multiple unmanned nodes to autonomously network and coordinate communication without the support of fixed infrastructure. It is widely used in scenarios such as military reconnaissance, disaster relief, and wide-area monitoring. This technology ensures network connectivity and data transmission reliability when nodes move or the topology changes through dynamic routing and distributed protocols.

[0003] Cluster-based unmanned self-organizing network systems represent a key technological direction for improving network scalability and management efficiency. The core idea is to divide the network nodes into multiple clusters, with each cluster electing a cluster head node responsible for intra-cluster coordination and inter-cluster communication, thereby forming a hierarchical network structure to reduce network control overhead and optimize resource allocation.

[0004] Static or semi-static clustering algorithms based on node ID, geographic location, or remaining energy are employed, and communication occurs on fixed or pre-allocated frequency bands. However, in complex battlefield or urban electromagnetic environments, available spectrum resources are dynamically changing and subject to significant interference, leading to low spectrum utilization efficiency in traditional fixed spectrum allocation models. Furthermore, existing clustering algorithms lack the ability to perceive and adapt to real-time electromagnetic environments, and their clustering structures and routing strategies fail to be optimized in tandem with dynamic spectrum conditions. This results in inter-cluster links being easily interrupted under interference, leading to a decrease in overall network throughput and an increase in latency. In addition, traditional centralized or periodic spectrum sensing methods are costly and lack real-time performance, failing to provide timely and accurate spectrum availability information for dynamic clustering and routing decisions, making it difficult to ensure continuous and reliable communication for highly dynamic unmanned swarms in complex electromagnetic environments. Summary of the Invention

[0005] The purpose of this invention is to provide a cluster-based unmanned cluster self-organizing network system to solve the problems of low spectrum utilization efficiency, poor inter-cluster link reliability, and overall network performance degradation caused by the disconnect between cluster structure, routing strategy, and spectrum resource status in complex dynamic electromagnetic environments.

[0006] This invention provides a cluster-based unmanned cluster self-organizing network system. This system is composed of homogeneous intelligent agent modules deployed on each unmanned node, forming a decentralized adaptive network. Specifically, the system includes: a distributed spectrum situational awareness module, a dynamic clustering and cluster head election module, a spectrum-topology joint optimization routing module, and a cross-layer collaborative control module. The distributed spectrum awareness module is used to detect and fuse local electromagnetic environment information in real time and collaboratively, generating a node-level spectrum availability map. The dynamic clustering and cluster head election module performs dynamic cluster structure partitioning and optimal cluster head election based on real-time environmental information and node status provided by the spectrum awareness module. The spectrum-topology joint optimization routing module calculates and selects the optimal communication path and channel for intra-cluster and inter-cluster communication based on the current cluster structure, cluster head status, and global spectrum situation. The cross-layer collaborative control module is responsible for coordinating the operation timing and data interaction of the above three modules, ensuring the overall consistency and real-time performance of system decisions.

[0007] Furthermore, the distributed spectrum situational awareness module operates as follows: Each unmanned node's sensing unit rapidly scans a preset wideband at millisecond intervals, collecting raw data on signal strength, noise floor, and occupancy status of multiple channels. The module incorporates a lightweight spectrum feature extraction algorithm to process the raw data, extracting the signal-to-interference-plus-noise ratio (SINR) estimate, expected idle time, and interference signal feature fingerprint for each channel. Subsequently, the node exchanges the processed local spectrum sensing results with its one-hop neighbor nodes via its ad hoc network communication link. The module also includes a distributed data fusion unit based on a consensus algorithm, used to weightedly fuse spectrum information from multiple neighbor nodes, ultimately generating a timestamped, refined spectrum availability map covering the communication radius of each node. This map is stored in the form of a data table, with entries including at least the channel number, comprehensive availability score, effective time window, and confidence index.

[0008] Furthermore, the execution of the dynamic clustering and cluster head election module is initiated by the cross-layer collaborative control module under the drive of a preset time-triggered event or network topology change event. The clustering algorithm of this module aims to maximize cluster stability and spectrum resource matching as a joint optimization objective. Specifically, the dynamic clustering and cluster head election module defines a multi-dimensional election weight value for each candidate cluster head. This weight value is calculated by linearly weighting the node's remaining energy factor, node mobility factor, and the spectrum quality factor of the node's location. The spectrum quality factor is directly derived from the average comprehensive availability score calculated for a preset set of inter-cluster communication backup channels in the node's spectrum availability map. This module exchanges the election weight values ​​of nodes within the local network through flooding and, according to a preset threshold comparison rule, allows the node with the highest weight value to automatically declare itself as the cluster head. Ordinary nodes within its communication range then choose to join the cluster, forming the initial cluster structure. For boundary nodes that fail to join any cluster, the module initiates a cluster merging or splitting negotiation process to ensure full network coverage.

[0009] Furthermore, as an embodiment of the present invention, the spectrum-topology joint optimization routing module includes an intra-cluster routing submodule and an inter-cluster routing submodule. The intra-cluster routing submodule employs a spectrum-aware on-demand distance vector improvement protocol. The route discovery process between ordinary nodes and cluster heads incorporates channel availability as a key metric into route request and response messages, prioritizing paths where the channel quality of each hop is higher than a threshold. The inter-cluster routing submodule maintains a virtual backbone topology with cluster heads as vertices. Each virtual link in the graph corresponds to a set of actually available physical links between a pair of cluster head nodes and its corresponding channel quality vector. When cross-cluster communication is required, this submodule first calculates an optimal virtual path based on the virtual backbone topology using a composite cost function that considers hop count, end-to-end delay, and the path's minimum spectrum availability score. Subsequently, for each virtual link on this virtual path, a suitable specific working channel is allocated in real-time from the set of available physical channels corresponding to that link, based on channel isolation principles and load balancing strategies, thereby instantiating the virtual path into a specific, spectrum-optimized end-to-end physical route.

[0010] Furthermore, the cross-layer collaborative control module achieves closed-loop coordination among clustering, routing, and spectrum awareness. This module defines and manages three system operating states: steady-state operation, local reconstruction, and global reconstruction. In steady-state operation, the system performs spectrum awareness and routing maintenance according to a basic cycle. When the distributed spectrum situational awareness module detects persistent strong interference in a local area, causing a significant degradation of the spectrum availability map of nodes in that area, the cross-layer collaborative control module triggers the local reconstruction state. In this state, the dynamic clustering and cluster head election modules within the affected local area are activated, re-electing cluster heads with better spectrum conditions and adjusting cluster membership relationships. Simultaneously, the spectrum-topology joint optimization routing module updates the affected routing table entries. When a large-scale topology change or a drastic change in the wide-area electromagnetic environment occurs, the cross-layer collaborative control module triggers the global reconstruction state, coordinating joint re-optimization of clustering and routing across the entire network.

[0011] Furthermore, the system employs an efficient channel access mechanism based on time slot allocation. This mechanism is coordinated by the cluster head node. The cluster head divides the superframe period into control time slot clusters and data time slot clusters based on the communication needs of its cluster members and the spectrum availability map. Within the control time slot cluster, all nodes use a pre-set, interference-resistant common control channel for signaling interaction, including the broadcasting of spectrum sensing results, and the transmission of clustering signaling and routing signaling. Within the data time slot cluster, the cluster head dynamically allocates specific data channels and time slots for communication between cluster members and between the cluster head itself. The allocated channels are selected from a pool of high-quality channels filtered from the current spectrum availability map. This mechanism achieves channel separation between the control plane and the data plane, ensuring reliable transmission of critical signaling while maximizing the freedom of spectrum utilization on the data plane.

[0012] Furthermore, the dynamic clustering and cluster head election module also introduces a cluster head load balancing mechanism. This mechanism sets a maximum and a minimum threshold for the number of members in a cluster head node. When the number of members in a cluster exceeds the maximum threshold, the cluster head will proactively initiate a cluster splitting process, dividing the cluster into two new clusters based on the geographical location and spectral characteristics of the member nodes. When the number of members in two adjacent clusters is below the minimum threshold, and the spectral compatibility between their cluster heads is good, the cross-layer collaborative control module will coordinate these two cluster heads to initiate a cluster merging process, elect a new cluster head, and merge them into one cluster to reduce management overhead.

[0013] Furthermore, when calculating inter-cluster virtual paths, the spectrum-topology joint optimization routing module uses a composite cost function that is: the total path cost equals the path hop count multiplied by a hop count weighting coefficient, plus the sum of the hop delays multiplied by a delay weighting coefficient, plus the sum of the reciprocals of the minimum spectrum availability scores for each hop multiplied by a spectrum weighting coefficient. The spectrum availability score is a normalized value between 0 and 1, with a higher score indicating better channel quality. Each weighting coefficient is dynamically adjusted by the cross-layer collaborative control module based on the current overall network service type and QoS requirements.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention, by introducing a distributed spectrum situational awareness module, achieves high-resolution, low-latency collaborative awareness of complex dynamic electromagnetic environments, providing a real-time and accurate spectrum situational information foundation for network decision-making. The awareness results directly drive subsequent clustering and routing optimization, enabling the network structure to proactively adapt to rather than passively withstand environmental changes, fundamentally solving the problem of blind decision-making caused by the lack or lag in environmental awareness in traditional solutions.

[0015] 2. This invention creatively integrates the real-time spectrum quality factor as a core parameter into the dynamic clustering and cluster head election algorithm, ensuring that the elected cluster heads are not only nodes with sufficient energy or central locations, but also nodes located in regions with superior spectrum conditions. This guarantees that the cluster head, a key network node, has a stable and reliable communication foundation, significantly improving the robustness of intra-cluster management signaling and inter-cluster relay links, thereby enhancing the stability and anti-interference capability of the entire hierarchical network structure.

[0016] 3. The spectrum-topology joint optimization routing mechanism proposed in this invention breaks the limitation of independent routing and spectrum allocation in traditional hierarchical networks. This mechanism simultaneously considers spectrum constraints during path optimization at the virtual backbone layer and performs refined channel allocation when finally instantiating the path. This cross-layer joint design enables data routing to proactively avoid interference areas and select paths with abundant spectrum resources, achieving optimal spatial matching between network topology and spectrum resources, effectively improving network throughput and reducing end-to-end transmission latency.

[0017] 4. This invention achieves intelligent closed-loop linkage between sensing, clustering, and routing through a cross-layer collaborative control module, forming a complete adaptive cycle of "sensing environmental changes, adjusting network structure, and optimizing transmission paths." The system can intelligently switch between various operating modes such as steady state, local reconstruction, and global reconstruction, responding to different levels of environmental and topology dynamics with varying resource overhead. This achieves an optimal balance between network performance and control overhead, ensuring the continuous, reliable, and efficient networking and communication capabilities of highly dynamic unmanned clusters in extremely complex electromagnetic environments. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall technical architecture of the unmanned cluster self-organizing network system based on clustering proposed in this invention; Figure 2 This is a schematic diagram illustrating the core principle framework of the spectrum-topology joint optimization routing mechanism in this invention; Figure 3 This is a logical flowchart of dynamic clustering and cluster head election in this invention; Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow of distributed spectrum situational awareness and cross-layer collaborative control in this invention; Figure 5 This is a schematic diagram illustrating the principle framework of the efficient channel access mechanism based on time slot allocation in this invention. Detailed Implementation

[0019] This invention provides a cluster-based unmanned self-organizing network system. Please refer to the appendix. Figures 1 to 5 This system comprises isomorphic intelligent agent modules deployed on each unmanned node, forming a decentralized adaptive network. The core of the system lies in the deep joint optimization of network structure, communication routing, and spectrum resource status through the close collaboration of multiple functional modules, thereby ensuring high performance and high reliability of the network in complex and dynamic electromagnetic environments. Specifically, the system includes a distributed spectrum situational awareness module, a dynamic clustering and cluster head election module, a spectrum-topology joint optimization routing module, and a cross-layer collaborative control module. These four modules are logically hierarchical and progressive, and operate in a closed-loop manner during runtime, jointly completing the entire adaptive process from environmental perception to network reconstruction and path optimization.

[0020] The distributed spectrum situational awareness module is the core unit of the system for sensing the external electromagnetic environment. Please refer to the attached document. Figure 4 This module operates independently on each unmanned node. Its core task is to detect and fuse local electromagnetic environment information in real time and collaboratively, ultimately generating a node-level, refined spectrum availability map for each node. The module's operation begins with its onboard hardware sensing unit. This sensing unit typically consists of a wideband RF front-end, a high-speed analog-to-digital converter, and a digital signal processor. It rapidly scans the entire preset operating bandwidth of the system at millisecond intervals, such as 10 or 20 milliseconds.

[0021] Within a single scan cycle, the sensing unit, according to a preset channel partitioning rule (e.g., dividing the 2.4 GHz to 2.4835 GHz frequency band into 15 channels with a width of 5 MHz), sequentially or in parallel collects raw data for each channel. This raw data includes at least three dimensions of information: signal strength, noise floor, and channel occupancy status. Signal strength, measured in decibels per milliwatt (dW), characterizes the presence and power level of external signals within the channel. Noise floor, also measured in DW, characterizes the background noise level of the channel. The channel occupancy status is a Boolean value, derived by the energy detector within the sensing unit by comparing the instantaneous signal power with a dynamically calculated energy detection threshold, used to preliminarily determine whether the channel is occupied.

[0022] After acquiring the raw data, the lightweight spectrum feature extraction algorithm built into the distributed spectrum situational awareness module immediately starts. This algorithm runs on the node's microprocessor, taking the aforementioned raw data sequence as input and outputting a multi-dimensional feature vector for each scanned channel. The feature extraction process comprises several sub-steps. The first step is signal-to-interference-plus-noise ratio (SINR) estimation. The algorithm first separates the possible useful signal components and interference noise components from the raw signal strength data. For each channel, the algorithm uses short-time Fourier transform to analyze the signal spectrum characteristics and, combined with noise baseline data, estimates an approximate SINR value. The SINR value is a dimensionless numerical value that preliminarily reflects the channel's communication quality potential. The second step is idle duration prediction. The algorithm analyzes the occupancy state sequence of the channel over multiple historical scanning cycles and, using a simplified hidden Markov model or time-series-based prediction method, estimates the expected idle time, measured in milliseconds, after the channel transitions from an occupancy state to an idle state. The third step is interference signal feature fingerprint extraction. For channels identified as occupied, the algorithm further analyzes their signal strength waveform, center frequency offset, bandwidth, and other characteristics to generate a simplified feature fingerprint vector. This fingerprint can be used to distinguish between different types of interference sources, such as periodic pulse interference from fixed radar and continuous wave interference from other communication systems.

[0023] After completing local feature extraction, nodes need to exchange processed local spectrum sensing results with neighboring nodes within their one-hop communication range via their ad hoc network communication links. To this end, the distributed spectrum situational awareness module defines a standardized spectrum information exchange message format. This message is encapsulated in the form of a data frame, with its header containing the source node identifier, sequence number, timestamp, and message type identifier. The message payload is a structured data table, where each row corresponds to a channel, and the columns include at least: channel number, estimated signal-to-interference-plus-noise ratio (SNR), estimated idle duration, interference feature fingerprint summary, and local confidence score. The local confidence score is a value between 0 and 1, calculated comprehensively based on factors such as the SNR of the sensing unit and the stability of the feature extraction algorithm, and is used to characterize the reliability of the sensing result.

[0024] When a node receives spectrum information exchange messages from one or more neighboring nodes, the distributed data fusion unit based on a consensus algorithm within the module is activated. The goal of this unit is to integrate the local observations of multiple nodes to generate a more comprehensive and accurate spectrum availability map. The fusion process is iterative. Each node maintains a local spectrum state table, initialized with its own perceived feature vectors. In each fusion iteration, the node broadcasts its local spectrum state table to all neighboring nodes and simultaneously receives the local spectrum state tables from all neighboring nodes. For each channel entry in the spectrum state table, the fusion algorithm performs a weighted average operation. Specifically, for continuous values ​​such as the signal-to-interference-plus-noise ratio (SINR) estimate and expected idle time, the fused value equals the node's value multiplied by its own confidence weight, plus the sum of all neighboring node values ​​multiplied by their respective confidence weights, and then divided by the sum of all weights.

[0025] The confidence weight is determined by the local confidence field in the message and the link quality index; the data weight of neighboring nodes with poor link quality is reduced. For discrete or vector data such as interference feature fingerprints, the fusion algorithm uses voting or clustering mechanisms to identify common interference patterns. After several rounds of iteration, the local spectrum state tables of all nodes will tend to be consistent, i.e., a distributed consensus is achieved. At this point, each node obtains a refined spectrum availability map with a unified timestamp covering its communication radius. This map is ultimately stored in the node's non-volatile memory in the form of a data table, with entries including at least the channel number, comprehensive availability score, effective time window, confidence index, and interference type identifier. The comprehensive availability score is a normalized scalar between 0 and 1, calculated from parameters such as the fused signal-to-interference-plus-noise ratio and expected idle time using a preset scoring function. A higher score indicates better overall channel quality at the current time and location. The effective time window predicts the effective duration of this score value.

[0026] The dynamic clustering and cluster head election modules are crucial for constructing a hierarchical network structure in the system. Please refer to the appendix. Figure 3 This module's execution is not periodic, but rather initiated by the cross-layer collaborative control module under the influence of preset specific events. These triggering events mainly fall into two categories: time-triggered events and network topology change events. Time-triggered events refer to the system reaching a preset stable period, such as actively initiating a clustering evaluation every 5 seconds. Network topology change events include new nodes joining the network, existing nodes leaving the network, or a drastic deterioration in local link quality detected through spectrum sensing. After the cross-layer collaborative control module issues a clustering command, the dynamic clustering and cluster head election module begins operation. The core algorithm of this module uses maximizing cluster stability and spectrum resource matching as a joint optimization objective. To achieve this objective, the module calculates a multi-dimensional election weight value for each candidate node in the network.

[0027] The calculation of the election weight value depends on several input parameters. The first parameter is the node's remaining energy factor. This factor is provided by the node's power management unit and is typically expressed as a percentage of the current remaining battery capacity to the initial full capacity, with a value between 0 and 100. High remaining energy means the node is capable of handling the additional computational and communication overhead required by the cluster head. The second parameter is the node mobility factor. This factor is estimated by the node's built-in GPS module or inertial measurement unit. The algorithm calculates the standard deviation of the node's position coordinates or average movement speed over a past time window, such as the last 3 seconds. Lower mobility results in a higher factor value, indicating a relatively stable node location, which is beneficial for the durability of the cluster structure. The third, and most critical, parameter is the spectral quality factor of the node's location. This factor is directly derived from the node's spectrum availability map. The module first filters out a pre-defined set of backup channels specifically for inter-cluster communication from the spectrum availability map. This set typically contains 3 to 5 channels, which are pre-configured as frequency bands with strong anti-interference characteristics. Then, the algorithm calculates the arithmetic mean of the overall availability scores of all channels in this set. This average value is the spectral quality factor of the node, which quantifies the quality of the spectral environment on which the node's external communication links depend when it acts as a cluster head.

[0028] After obtaining the above three factors, the dynamic clustering and cluster head election module calculates the final election weight value for each node using a linear weighted formula. This formula is: Election weight value equals Remaining energy factor × Energy weight coefficient + Mobility factor × Mobility weight coefficient + Spectrum quality factor × Spectrum weight coefficient. Here, the energy weight coefficient, mobility weight coefficient, and spectrum weight coefficient are a set of preset normalized coefficients, and their sum is 1. The specific values ​​of these coefficients are dynamically fine-tuned by the cross-layer collaborative control module based on the overall network service load and anti-interference requirements. For example, the spectrum weight coefficient will be increased when the electromagnetic environment is severe; the energy weight coefficient will be increased when network energy is generally insufficient.

[0029] After calculating the election weights, the module exchanges this information within the local network using restricted flooding. Each node constructs a signaling message containing its own node identifier, election weight, geographical location, and current state, and broadcasts it within its communication range. Simultaneously, nodes also receive similar messages from neighboring nodes. Each node processes this information according to a preset threshold comparison rule. A typical rule is that a node compares its own election weight with the election weights of all its neighboring nodes. If its own weight consistently exceeds that of all neighboring nodes within the comparison time window and exceeds a preset relative threshold, such as exceeding the second-highest weight node by 10%, then the node automatically declares itself a cluster head. The announcement signaling is broadcast. Ordinary nodes receiving the cluster head announcement signaling, if not already joined any cluster, will send a join request to the cluster head and record that the cluster head is their assigned cluster head, thus forming the initial cluster structure. The cluster head maintains a cluster member list, recording the identifiers and link quality of all joining nodes.

[0030] For isolated or boundary nodes that fail to join any cluster, the dynamic clustering and cluster head election module initiates a negotiation process for cluster merging or splitting. This negotiation process is coordinated and initiated by the cross-layer collaborative control module. For nodes with weak signals due to being at the edge of multiple clusters, they simultaneously listen to signaling from multiple cluster heads and attempt to join the cluster with the strongest signal and the highest cluster head election weight. If joining fails, the node reports its status, and the cross-layer collaborative control module instructs its nearest cluster head to appropriately increase communication power or adjust the cluster boundary to incorporate it. For excessively small clusters formed due to low node density, the module assesses the feasibility of merging adjacent clusters.

[0031] The spectrum-topology joint optimization routing module is the core of the system for achieving efficient and reliable data transmission. Please refer to the attached document. Figure 2 This module calculates and selects the optimal communication path and channel for intra-cluster and inter-cluster communication based on the current cluster structure, cluster head status, and global spectrum situation. Logically, this module contains two relatively independent but closely coordinated sub-modules: an intra-cluster routing sub-module and an inter-cluster routing sub-module.

[0032] The intra-cluster routing submodule manages the communication paths between all ordinary nodes within the cluster and the cluster head node. This submodule employs a spectrum-aware, on-demand distance vector improved protocol. When an ordinary node within the cluster needs to send data to the cluster head or needs to communicate with other clusters via the cluster head relay, it will initiate a route discovery process if there is no valid path in its routing table. This node generates a route request message. Unlike traditional on-demand distance vector protocols, the route request message of this invention adds a "path spectrum quality record" field to the original fields. When an intermediate node forwards this route request message, it not only records the previous hop address and hop count, but must also query the local spectrum availability map to obtain the comprehensive availability score of the channel recommended for the link from the previous hop node to itself, and append this score value to the "path spectrum quality record" field. This field essentially forms a vector of the channel quality for each hop along a path. When the route request message finally reaches the destination node, the cluster head, the cluster head may receive multiple copies of the route request message from different paths.

[0033] The intra-cluster routing submodule at the cluster head evaluates these paths. The evaluation rule is as follows: First, it checks all score values ​​in the "Path Spectrum Quality Record" field for each path. If any score is below a preset threshold, such as 0.6, the path is discarded. Then, among the remaining paths, the path with the fewest hops and the highest spectrum score is selected as the optimal path. Subsequently, the cluster head sends a routing response message in reverse along the selected optimal path, and nodes along the way establish forward routing table entries. The routing table entries contain not only the next-hop address but also the channel number suggested for that hop.

[0034] The inter-cluster routing submodule manages cross-cluster communication on the virtual backbone with cluster heads as nodes. The core of this submodule is maintaining the virtual backbone topology graph. Each vertex in the graph represents a cluster head node. A virtual edge exists between two vertices if and only if there is at least one available direct physical radio link between the two cluster head nodes, and the two clusters do not belong to the same merge candidate pair. For each virtual edge, the inter-cluster routing submodule provides a set of associated physical link information. This set contains all available physical channels between the pair of cluster head nodes and their corresponding quality information. This information originates from the intersection of the spectrum availability maps of the two cluster head nodes and is synchronously updated through periodic signaling exchanges between the cluster heads. Therefore, each virtual edge actually corresponds to a list of candidate channels and the quality vector of each channel.

[0035] When cross-cluster communication is required, such as when a regular node in cluster A needs to communicate with a regular node in cluster B, the data is first aggregated to its respective cluster head via intra-cluster routing. Subsequently, the inter-cluster routing submodule on the source cluster head A is triggered. This submodule first calculates one or more virtual paths from the source cluster head A to the destination cluster head B using a composite cost function based on the virtual backbone topology. This composite cost function is the mathematical embodiment of the spectrum-topology joint optimization of this invention. Specifically, for a virtual path, its total cost equals the hop count multiplied by the hop count weight coefficient, plus the sum of the estimated transmission delays of each virtual link segment on the path multiplied by the delay weight coefficient, plus the sum of the reciprocals of the minimum spectrum availability scores corresponding to each virtual link segment on the path multiplied by the spectrum weight coefficient.

[0036] ; In this formula, This represents the total cost of the path; the lower the cost, the better the path. This represents the number of hops in the path. Representing the The estimated transmission delay of the hop virtual link is estimated from the link distance, modulation scheme, and channel bandwidth. Representing the The highest overall availability score among all available channels in the set of physical link information associated with the hop virtual link is the best spectrum quality that the hop link can obtain. , , These are the hop count weighting coefficient, latency weighting coefficient, and spectrum weighting coefficient. These coefficients are dynamically adjusted by the cross-layer collaborative control module based on the overall network service type and quality of service requirements. For example, when the service is extremely sensitive to latency... The value increases; when the electromagnetic environment is complex and the requirements for link reliability are high, The value increases. Spectrum availability score It is a normalized value between 0 and 1. A higher score indicates better channel quality, and its reciprocal is 1 / The smaller the value, the smaller its contribution to the total cost, which motivates the algorithm to select links with high spectrum quality.

[0037] The inter-cluster routing submodule uses an improved Dijkstra's algorithm or the A* algorithm, based on the aforementioned composite cost function, to calculate a virtual path with the minimum total cost in the virtual backbone network topology. The task does not end after calculating the virtual path, as each hop on the virtual path is still only a logical connection. Next, path instantiation is required, i.e., assigning a suitable specific working channel to each hop virtual link. The instantiation process is completed collaboratively by the cluster heads on the path. For the first hop on the virtual path, i.e., from source cluster head A to the next relay cluster head C, source cluster head A selects the final channel from its set of physical link information associated with cluster head C, based on the channel isolation principle and load balancing strategy. The channel isolation principle requires that the selected channel be as far away as possible in the frequency domain from other channels currently used by the node, especially its intra-cluster communication channels, to reduce self-interference. The load balancing strategy considers the channel's occupancy in adjacent areas, prioritizing channels with lighter loads. Source cluster head A notifies the next-hop cluster head C of the selected channel number via the control channel. Cluster head C confirms that the channel is also available and of acceptable quality in its local spectrum availability map, and then replies with confirmation. Cluster head C then performs the same channel selection procedure for the next hop of the path, proceeding hop by hop until reaching the destination cluster head B. At this point, a specific, spectrum-optimized end-to-end physical route has been successfully established. Each hop of this route specifies the sending node, the receiving node, and the specific channel used.

[0038] The cross-layer collaborative control module is the nerve center of the entire system, responsible for coordinating the operational timing and data interaction of the distributed spectrum situational awareness module, the dynamic clustering and cluster head election module, and the spectrum-topology joint optimization routing module, ensuring the overall consistency and real-time performance of system decisions. Please refer to the appendix again. Figure 4 This module achieves closed-loop coordination among clustering, routing, and spectrum sensing. To this end, the cross-layer collaborative control module defines and manages three system operating states: steady-state operation, local reconfiguration, and global reconfiguration. Each state corresponds to a set of preset module execution strategies and parameter configurations.

[0039] In steady-state operation, the network topology and electromagnetic environment are relatively stable. At this time, the cross-layer collaborative control module coordinates the system to execute routine tasks according to the basic cycle. The distributed spectrum situation awareness module operates at a higher frequency, for example, performing a fast scan and lightweight synchronization with neighbor information every 50 milliseconds to maintain the timeliness of the spectrum availability map. The spectrum-topology joint optimization routing module maintains the routing table as needed and periodically sends keep-alive messages to detect link connectivity. The dynamic clustering and cluster head election module is in a dormant or low-power listening state and does not actively initiate clustering operations. In this state, the control module mainly monitors two indicators: the local rate of change of the spectrum situation and the alarm frequency of routing link failures.

[0040] When the distributed spectrum situational awareness module detects persistent strong interference in a local area and determines through data fusion that the spectrum availability map of multiple nodes in that area has significantly deteriorated (e.g., more than 30% of the critical channel scores drop by more than 50% over three consecutive cycles), it sends a local environmental degradation event report to the cross-layer collaborative control module. Upon receiving this report, the cross-layer collaborative control module switches the system state from steady-state operation to a local reconstruction state. In this state, the control module first delineates the affected area, typically all nodes within a two-hop communication radius centered on the event reporting node. Then, it sends instructions to all nodes within this area to activate their dynamic clustering and cluster head election modules. At the same time starting point coordinated by the control module, these nodes recalculate election weights based on the latest, deteriorated spectrum availability map and execute the cluster head election process.

[0041] Due to a significant drop in the spectrum quality factor, the election weight of the original cluster head node is likely to decrease, potentially leading to the election of a new cluster head located at the edge of the affected area with relatively better spectrum conditions. After the new cluster head is elected, cluster membership is readjusted. Simultaneously, the control module notifies the spectrum-topology joint optimization routing module that routing table entries in this area may be invalid and require synchronous updates. The routing module then instructs nodes associated with the invalidated paths to initiate a new route discovery process. The entire local reconstruction process is confined to the affected local area, avoiding network-wide control overhead.

[0042] When a large-scale network topology change or a drastic change in the wide-area electromagnetic environment occurs, the cross-layer collaborative control module triggers a global reconfiguration state. Large-scale topology changes may be caused by unmanned clusters performing formation changes, collective failures of some nodes, or the addition of new nodes. Dramatic changes in the wide-area electromagnetic environment may be caused by full-band suppression interference or natural environmental factors. Criteria for triggering global reconfiguration include: a certain proportion of nodes simultaneously reporting link outages, or the network-wide average spectrum availability score plummeting below a dangerous threshold within a short period. Upon entering the global reconfiguration state, the cross-layer collaborative control module broadcasts a reconfiguration command to the entire network. The distributed spectrum situational awareness module is required to immediately perform a complete, high-precision collaborative sensing and data fusion to obtain the network's baseline spectrum situation. Subsequently, the dynamic clustering and cluster head election module re-executes clustering and cluster head election across the entire network, constructing the optimal cluster head based on the new global environmental information.

Claims

1. A cluster-based unmanned self-organizing network system, characterized in that, include: The distributed spectrum situational awareness module is used for real-time collaborative detection and fusion of local electromagnetic environment information to generate a node-level spectrum availability map. The dynamic clustering and cluster head election module is used to perform dynamic cluster structure partitioning and optimal cluster head election based on the spectrum availability map and node status provided by the distributed spectrum situation awareness module. The spectrum-topology joint optimization routing module is used to calculate and select the optimal communication path and channel for intra-cluster and inter-cluster communication based on the current cluster structure, cluster head status and global spectrum situation. The cross-layer collaborative control module is used to coordinate the operation timing and data interaction of the distributed spectrum situational awareness module, the dynamic clustering and cluster head election module, and the spectrum-topology joint optimization routing module, to ensure the overall consistency and real-time performance of system decisions.

2. The cluster-based unmanned self-organizing network system according to claim 1, characterized in that, The distributed spectrum situation awareness module includes a sensing unit, a spectrum feature extraction algorithm, and a distributed data fusion unit. The sensing unit performs a rapid scan of a preset wideband at millisecond intervals to collect raw data on signal strength, noise floor, and occupancy status of multiple channels. The spectrum feature extraction algorithm processes the raw data to extract the signal-to-interference-plus-noise ratio estimate, expected idle duration, and interference signal feature fingerprint for each channel. The distributed data fusion unit, based on a consensus algorithm, performs weighted fusion of spectrum information from multiple neighboring nodes to generate a refined spectrum availability map covering the communication radius of the covered nodes. The spectrum availability map is stored in the form of a data table, and its entries include at least the channel number, comprehensive availability score, effective time window, and confidence index.

3. The cluster-based unmanned self-organizing network system according to claim 2, characterized in that, The dynamic clustering and cluster head election module calculates a multidimensional election weight value for each candidate cluster head; The multidimensional election weight value is calculated by linearly weighting the node's remaining energy factor, node mobility factor, and the spectral quality factor of the node's location; The spectrum quality factor is derived from the average comprehensive availability score calculated for a preset set of inter-cluster communication backup channels in the spectrum availability map of the node. The dynamic clustering and cluster head election module exchanges the election weight values ​​of nodes through flooding and, according to a preset threshold comparison rule, makes the node with the highest weight value declare itself as the cluster head, and ordinary nodes within its communication range join the cluster to form the initial cluster structure.

4. The cluster-based unmanned self-organizing network system according to claim 3, characterized in that, The spectrum-topology joint optimization routing module includes an intra-cluster routing submodule and an inter-cluster routing submodule; The intra-cluster routing submodule adopts a spectrum-aware on-demand distance vector improvement protocol. During the route discovery process, channel availability is incorporated as a key metric into the route request and route response messages, and paths in which the quality of each hop channel is higher than the threshold value are given priority. The inter-cluster routing submodule maintains a virtual backbone network topology with cluster heads as vertices. Each virtual link in the graph corresponds to a set of actually available physical links between a pair of cluster head nodes and their corresponding channel quality vector. The inter-cluster routing submodule calculates the optimal virtual path based on the virtual backbone network topology using a composite cost function that considers hop count, end-to-end latency, and path minimum spectrum availability score. For each virtual link on the virtual path, the inter-cluster routing submodule allocates a suitable specific working channel in real time from the set of available physical channels corresponding to that link, based on the channel isolation principle and load balancing strategy, thereby instantiating the virtual path into a spectrum-optimized end-to-end physical route.

5. The cluster-based unmanned self-organizing network system according to claim 4, characterized in that, The cross-layer collaborative control module defines and manages three system operating states: steady-state operation, local reconstruction, and global reconstruction. Under steady-state operation, the system performs spectrum sensing and route maintenance according to the basic cycle; When the distributed spectrum situational awareness module detects that a persistent strong interference in a local area has caused a significant deterioration of the spectrum availability map, the cross-layer collaborative control module triggers a local reconstruction state, activates the dynamic clustering and cluster head election module in the affected area to re-elect cluster heads and adjust cluster membership relationships, and simultaneously instructs the spectrum-topology joint optimization routing module to synchronously update the affected routing table entries. When a large-scale topology change occurs in the network or a drastic change occurs in the wide-area electromagnetic environment, the cross-layer collaborative control module triggers a global reconstruction state, coordinating the clustering and routing across the entire network for joint re-optimization.

6. The cluster-based unmanned self-organizing network system according to claim 5, characterized in that, The specific form of the composite cost function is as follows: the total path cost equals the number of hops multiplied by the hop count weight coefficient, plus the sum of the delays of each hop multiplied by the delay weight coefficient, plus the sum of the reciprocals of the minimum spectrum availability scores of each hop multiplied by the spectrum weight coefficient; where the spectrum availability score is a normalized value between 0 and 1, and a higher score indicates better channel quality; the hop count weight coefficient, delay weight coefficient, and spectrum weight coefficient are dynamically adjusted by the cross-layer collaborative control module according to the current overall network service type and service quality requirements.

7. The cluster-based unmanned self-organizing network system according to claim 6, characterized in that, The system also includes an efficient channel access mechanism based on time slot allocation; the efficient channel access mechanism is coordinated by the cluster head node; the cluster head divides the superframe period into control time slot clusters and data time slot clusters according to the communication needs of its members and the spectrum availability map; within the control time slot cluster, all nodes use a preset, interference-resistant common control channel for signaling interaction; within the data time slot cluster, the cluster head dynamically allocates specific data channels and time slots for communication between members within the cluster and for inter-cluster communication within the cluster head itself; the allocated data channels are all selected from a pool of high-quality channels filtered from the current spectrum availability map.

8. The cluster-based unmanned self-organizing network system according to claim 7, characterized in that, The dynamic clustering and cluster head election module also includes a cluster head load balancing mechanism. The cluster head load balancing mechanism sets a maximum and minimum threshold for the number of members in a cluster head node. When the number of members in a cluster exceeds the maximum threshold, the cluster head actively initiates a cluster splitting process, dividing the cluster into two new clusters based on the geographical location and spectral characteristics of the member nodes. When the number of members in two adjacent clusters is below the minimum threshold and the spectral compatibility between their cluster heads is good, the cross-layer collaborative control module coordinates these two cluster heads to initiate a cluster merging process, elect a new cluster head, and merge them into one cluster.

9. The cluster-based unmanned self-organizing network system according to claim 8, characterized in that, The execution of the dynamic clustering and cluster head election module is initiated by the cross-layer collaborative control module under the drive of a preset time-triggered event or a network topology change event; the time-triggered event is when the system reaches a preset stable period; the network topology change event includes a new node joining the network, an existing node leaving the network, or a severe deterioration in the quality of a local link detected by spectrum sensing.

10. The cluster-based unmanned self-organizing network system according to claim 9, characterized in that, The intra-cluster routing submodule of the spectrum-topology joint optimization routing module includes a path spectrum quality record field in its routing request message; when an intermediate node forwards a routing request message, it queries the local spectrum availability map, obtains the comprehensive availability score of the suggested channel for the link from the previous hop node to itself, and appends the score value to the path spectrum quality record field. The destination node cluster head selects the optimal path based on whether the score values ​​in the path spectrum quality record field are all higher than the preset threshold value and the path hop count.

Images