A fixed-wing target drone cluster control method with anti-destroying topology and self-reconfiguration capability

Abstract

The application discloses a fixed-wing target drone cluster control method with anti-destroying topology and self-reconfiguration capability, and belongs to the technical field of unmanned aerial vehicle cluster control. The method comprises the following steps: firstly, a k-connected anti-destroying topology is constructed, and the cluster connectivity is checked through algebraic connectivity; then, distributed consistency control is adopted to make the cluster speed and position tend to be consistent and adapt to the fixed-wing non-complete constraint; in combination with double detection of the physical layer and the state layer, the node failure is confirmed through local voting of neighbor nodes; after the failure, the topology self-reconfiguration is triggered immediately, the adjacency matrix and the Laplacian matrix are updated, the k-connected structure is recovered and the formation gap is filled; finally, a distributed task auction algorithm is adopted to redistribute the unfinished tasks of the failed nodes within the range of 2-hop neighbors. The application adopts a decentralized architecture, does not need a central node, can tolerate multiple node damages, and realizes layered linkage of topology reconfiguration and task redistribution, so that the formation stability and task continuity can be maintained under a complex electromagnetic interference environment, and the anti-destroying survival capability and the task completion rate of the target drone cluster are significantly improved.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) swarm control and intelligent collaboration technology, specifically relating to a fixed-wing target drone swarm control method with a damage-resistant topology and mission self-reconfiguration capability under complex electromagnetic interference and partial node damage conditions. Background Technology

[0002] Existing fixed-wing target drones typically fly along preset routes or use simple formation control methods. When the cluster is subjected to electromagnetic interference, communication interruption, or individual nodes are shot down in a combat training environment, the traditional centralized control structure is prone to problems such as formation instability, mission interruption, cluster disintegration, and the inability of remaining nodes to coordinate.

[0003] Currently, most control strategies rely on a single central node or a fixed communication topology. Once the core node is damaged, the control performance of the entire cluster will be severely degraded.

[0004] Meanwhile, existing anti-damage control solutions for UAV swarms are mostly geared towards multi-rotor UAV scenarios and do not fully consider the non-holonomic dynamic constraints (minimum turning radius, nominal airspeed range, maximum maneuver overload limit) of fixed-wing target drones, making them unsuitable for high-speed formation flight scenarios of fixed-wing target drones. Furthermore, existing solutions often design topology reconstruction and task redistribution separately. Topology adjustments do not consider mission execution requirements, and task redistribution does not match topology anti-damage constraints. In complex adversarial training environments, this can easily lead to a disconnect between formation reconstruction and mission execution, further exacerbating the risk of swarm instability and mission interruption.

[0005] Therefore, it is necessary to propose a control method that can simultaneously satisfy decentralization, topology resilience, support adaptive reconstruction after node damage, and maintain the continuity of formation and tasks. Summary of the Invention

[0006] To address the aforementioned shortcomings of existing fixed-wing target drone swarm control technologies, this invention aims to propose a fixed-wing target drone swarm control method based on distributed graph topology control and a mission self-reconfiguration mechanism. This method solves the problems of formation instability, mission interruption, and swarm disintegration in traditional centralized control structures under scenarios of node damage and communication failure. Simultaneously, it adapts to the flight dynamics characteristics of fixed-wing target drones, achieving deep linkage between topology damage-resistant reconfiguration and adaptive mission reallocation. Even with partial node damage or communication link failure, it can still stably maintain the formation structure and mission continuity.

[0007] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0008] A method for controlling a fixed-wing target drone swarm with survivable topology and self-reconfiguration capabilities includes the following steps:

[0009] 1) Construct a robust topology, abstracting the fixed-wing target drone cluster into an undirected graph model, and simultaneously constructing... - A connected graph structure ensures that each node in the cluster is connected to at least one other node. Each neighboring node maintains a communication link, and the cluster connectivity is verified by the algebraic connectivity of the Laplace matrix;

[0010] 2) Distributed consensus control: Local consensus protocol is used for grouping control of each target machine;

[0011] 3) A dual verification architecture of physical layer communication detection and state layer flight rationality detection is adopted for node damage detection. When a node abnormality is detected, the final determination of node failure is completed through the local voting mechanism of neighboring nodes.

[0012] 4) Once a node failure is confirmed, perform topology self-reconstruction, update the adjacency matrix and Laplace matrix, re-verify the algebraic connectivity, and restore the topology. - Connect the structure and synchronously fill the formation gaps using the principle of minimum displacement adjustment;

[0013] 5) A distributed task auction algorithm is used for adaptive task reallocation. Only the unfinished tasks of failed nodes are reallocated within their 2-hop neighbors. The nodes participating in the auction select the task takeover node based on the cost function.

[0014] In the survivability topology, the nodes of the undirected graph model are individual target machines, and the edges are bidirectional communication links between nodes. The construction process is as follows:

[0015] After the cluster powers on, it performs initial neighbor discovery, generating an initial adjacency matrix based on communication distance, link signal-to-noise ratio, and relative flight stability. Based on adjacency matrix Calculate the degree matrix This leads to the generation of the Laplace matrix. Based on the Laplace matrix Perform topology pre-validation to confirm whether the initial topology meets the requirements. If the connectivity requirement is not met, the adjacent edges are re-optimized until the verification is passed.

[0016] The - Connected graph structure The value represents the maximum number of failed nodes the cluster can tolerate simultaneously, in relation to the cluster size. Strong binding, when the cluster size hour ,when hour ,and Meanwhile, the topology construction also satisfies the hard constraints of the fixed-wing target drone.

[0017] The adjacency matrix Specifically:

[0018]

[0019] in, Representation matrix The Line 1 Column elements, , They represent the first frame and the first Spatial coordinates of the target launcher Indicates communication distance. This represents the Euclidean distance between the two target drones. Indicates the first frame and the first Signal-to-noise ratio of the link between target drones This indicates the preset signal-to-noise ratio threshold.

[0020] The method of verifying cluster connectivity using the algebraic connectivity of the Laplace matrix specifically involves:

[0021]

[0022] Where L is the Laplacian matrix corresponding to the clustered undirected graph. It is the second smallest eigenvalue of the Laplace matrix, i.e., the algebraic connectivity.

[0023] The construction of the undisturbed topology structure satisfies the following three supporting mechanisms:

[0024] Multiple redundant communication paths: ensuring that at least one communication path exists between any two nodes within the cluster. Disjoint communication paths, with path priority based on link signal-to-noise ratio. Number of jumps Long-term stability ranking; automatically switches to a backup path when a single link fails.

[0025] Local Neighborhood Backup Node Mechanism: A backup node is selected for each node, based on the following criteria: the node with the best link quality, the highest remaining energy, and the most stable flight status among its neighbors. 1 node The backup node synchronizes the flight status, mission parameters, and formation information of the master node in real time, and has the ability to quickly take over after the master node fails.

[0026] Decentralized collaborative decision-making: Employing a local consensus mechanism, cluster decisions can be completed with only a majority vote from the neighboring nodes within two hops.

[0027] The control law of the local consensus protocol is:

[0028]

[0029] in: For the first Control inputs for each target node For the first Each target machine state vector These correspond to the node's three-dimensional spatial position, airspeed, track inclination, and heading angle, respectively. Gather for the neighbors, Adjacency matrix The corresponding adjacency weights.

[0030] Step 3) specifically refers to:

[0031] 3.1) When a node is detected to have experienced a communication timeout, abnormal status, or lost heartbeat signal, an anomaly warning is triggered and a misjudgment avoidance mechanism is executed. That is, when an abnormal node is detected, a query frame is first sent to the target node, and the backup node pre-synchronization process of the target node is started. The node is not directly judged as failed. Only when multiple neighboring nodes detect that the target node is abnormal will a local voting judgment mechanism be started to confirm the failed or damaged node. Specifically, the local voting judgment mechanism is as follows: within the two-hop neighbor range of the failed node, more than 50% of the valid nodes vote to judge the node as failed, that is, the node is finally confirmed as damaged or failed.

[0032] 3.2) After a node failure is confirmed, neighboring nodes immediately freeze the adjacency weight of the node, stop information exchange, broadcast the failed node ID to the range of 2-hop neighbors, and trigger the topology self-reconstruction process.

[0033] Step 4) specifically refers to:

[0034] 4.1) Update the adjacency matrix : Delete the row and column corresponding to the failed node, and add new adjacent edges to the neighboring nodes of the original failed node to ensure that the number of neighbors of all valid nodes is equal. ,satisfy The underlying requirements for connectivity;

[0035] 4.2) Reconstructing the Laplace matrix Based on the updated adjacency matrix Calculate the new degree matrix Generate the reconstructed Laplace matrix Immediately calculate the new algebraic connectivity. And verify If the connectivity requirement is not met, continue to add adjacent edges until the connectivity check passes.

[0036] 4.3) Recalculate the local neighbor set The decentralized update rule is adopted, so that nodes only interact with information with their one-hop neighbors;

[0037] 4.4) Automatic recovery Connectivity structure: when all nodes have degree , The topology reconstruction is considered complete when the topology remains stable for two consecutive control cycles.

[0038] The distributed task auction algorithm is as follows:

[0039] Auction Initiation: The original backup node of the failed node acts as the initiator, broadcasting the failed node's unfinished task information, task deadline, and execution requirements to its 2-hop neighbors.

[0040] Bidding price: Calculate the cost of nodes that meet the task execution requirements. The bid information is broadcast to the initiator within a preset time window;

[0041] Auction Decision: After receiving all bids, the initiating party selects the price. The smallest node acts as the task taker and broadcasts the winning bid results to the bidding area;

[0042] Task handover: The winning node synchronizes all task parameters, updates its own task queue, broadcasts a task takeover confirmation signal to the entire cluster, and completes task redistribution.

[0043] The task execution cost function is as follows:

[0044]

[0045] in: The distance cost from the node to the task execution point. The remaining energy cost for a node to perform a task. , The weighting coefficients for distance cost and energy cost are respectively, satisfying... + ;

[0046] The cost function undergoes dimensional normalization, reducing the distance cost. Energy cost Unified mapping to For multi-payload target scenarios, the extended cost function will be... ,in, For the cost of mission capabilities, This is the weighting coefficient for the cost of task capability.

[0047] The present invention has the following beneficial effects and advantages:

[0048] 1. Fully decentralized architecture, no central control node required, avoiding overall cluster loss of control due to core node failure.

[0049] 2. Based on the k-connected graph, the robust topology design supports maintaining global cluster connectivity even when multiple nodes are damaged, and can maintain stable formation when the number of damaged nodes is within 30%.

[0050] 3. Topology reconstruction and formation control are deeply integrated. After a node fails, topology self-repair and formation filling can be completed quickly, and the formation error convergence time is less than 5 seconds.

[0051] 4. Topology reconstruction and task redistribution are linked in a layered manner to ensure the continuity of task execution, with a task completion rate of ≥90% in complex adversarial scenarios.

[0052] 5. It fully adapts to the nonholonomic dynamic constraints of fixed-wing target drones, and adopts local communication and local decision-making mechanisms to significantly reduce communication overhead, making it suitable for combat training environments with complex electromagnetic interference. Attached Figure Description

[0053] Figure 1 System overall structure schematic diagram;

[0054] Figure 2 Damage-resistant topology and self-reconfiguration flowchart;

[0055] Figure 3 Single-node control structure block diagram;

[0056] Figure 4 Comparison of cluster k-connected topology structures;

[0057] Figure 5 Distributed task auction flowchart. Detailed Implementation

[0058] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0059] like Figures 1-3 As shown, a method for controlling a fixed-wing target drone swarm with survivable topology and self-reconfiguration capabilities includes the following steps:

[0060] Step 1: Construct a survivable topology

[0061] A fixed-wing target drone swarm is abstracted as an undirected graph model, where nodes correspond to individual target drones and edges represent bidirectional communication links between nodes. Based on the flight dynamics constraints and swarm resilience requirements of fixed-wing target drones, a [model / structure] is constructed. Connected graph structure, such as Figure 4 As shown, this ensures that each node in the cluster has at least one connection with... Each neighboring node maintains a stable communication link.

[0062] In this step, The rules for quantifying and selecting values ​​are as follows: The value represents the maximum number of failed nodes the cluster can tolerate simultaneously, in relation to the cluster size. Strong binding, when the cluster size hour ,when hour ,and Meanwhile, the topology construction must meet the hard constraints of the fixed-wing target drone, including minimum turning radius, maximum roll angle, nominal airspeed range, and minimum safe flight distance, to avoid the topology design exceeding the flight dynamics boundary.

[0063] Cluster connectivity is verified using the algebraic connectivity of the Laplace matrix, with the following criterion:

[0064]

[0065] Where L is the Laplacian matrix corresponding to the clustered undirected graph. It is the second smallest eigenvalue of the Laplace matrix, i.e., the algebraic connectivity.

[0066] The complete execution flow of topology construction is as follows: after the cluster is powered on, initial neighbor discovery is performed (beacon broadcast mechanism, matching of initial heartbeat period and control period) → an initial adjacency matrix is ​​generated based on communication distance, link signal-to-noise ratio, and relative flight stability. Based on adjacency moments Calculate the degree matrix Generate the Laplace matrix Topology pre-validation confirms that the initial topology satisfies Connectivity requirement and If the value is >0, the adjacent edges are re-optimized until the verification is passed.

[0067]

[0068] in, Representation matrix The Line number Column elements, , Representing the first frame and the first Spatial coordinates of the target launcher Indicates communication distance. This represents the Euclidean distance between the two target drones. Indicates the first frame and the first Signal-to-noise ratio of the link between target drones This indicates the preset signal-to-noise ratio threshold.

[0069] In a preferred embodiment of the present invention, if the distance between two nodes is less than the communication radius and the link quality meets the standard, then Otherwise .

[0070] Degree matrix It is a diagonal matrix, and its calculation is based entirely on the adjacency matrix. .

[0071]

[0072] Among them, the elements on the diagonal (i.e., node) The formula for calculating the degree is:

[0073]

[0074] Indicates the first The degree of a node, that is, the number of its neighbors. This indicates the total number of target machines in the cluster.

[0075] To further enhance topology resilience, three supporting mechanisms were designed simultaneously:

[0076] 1. Multiple redundant communication paths: ensuring that there is at least one communication path between any two nodes in the cluster. Disjoint communication paths, with path priority based on link signal-to-noise ratio. Number of jumps Long-term stability ranking; automatically switches to backup path when a single link fails.

[0077] 2. Local Neighborhood Backup Node Mechanism: A backup node is selected for each node, based on the following criteria: the node with the best link quality, the highest remaining energy, and the most stable flight status among its neighbors. 1 node The backup node synchronizes the flight status, mission parameters, and formation information of the master node in real time, and has the ability to quickly take over after the master node fails.

[0078] 3. Decentralized collaborative decision-making: Adopting a local consensus mechanism, cluster decisions can be completed by voting from a majority of nodes within two hops of neighbors, without the need for global communication and central node management, making it suitable for local communication scenarios in complex electromagnetic environments.

[0079] This step builds The interconnected and resilient topology provides underlying connectivity assurance for subsequent formation control. Step 1 completes the initial topology construction and passes the connectivity verification before step 2, distributed consistency control, can be initiated. During the operation of step 2, the topology parameters of this step are continuously reused, and the node status synchronization data is fed back to the topology evaluation module in real time to continuously verify the cluster connectivity, forming a real-time closed loop of "topology design - formation control".

[0080] Step 2: Distributed Consistency Control

[0081] Each target device employs a local consensus protocol, and the control law expression is as follows:

[0082]

[0083] in: For the first The control inputs of each target drone node are matched with the control deflection angle and throttle control quantity execution boundaries of the fixed-wing flight control system. For the first A target drone state vector, adapted to fixed-wing nonholonomic constraints, is defined as follows: These correspond to the node's three-dimensional spatial position, airspeed, track inclination, and heading angle, respectively. For the neighbor set, reuse the topology parameters constructed in step one; Adjacency matrix The corresponding adjacency weights are adjusted so that the weights decrease adaptively when the link quality deteriorates and are restored to their original values ​​when the link is stable, thus avoiding the impact of weak communication links on the formation convergence.

[0084] This step employs a cascaded consistency control architecture, divided into a speed consistency inner loop and a position consistency outer loop: the speed consistency inner loop ensures that the airspeed, heading, and altitude change rates of nodes within the cluster are consistent, adapting to the inner loop control logic of fixed-wing flight control; the position consistency outer loop ensures that the relative positions of the formation converge to the preset formation, achieving stable formation maintenance. Based on Lyapunov stability theory, it can be proven that when the cluster satisfies... When the connectivity criterion is used, this control law can achieve global asymptotic consistency of the cluster state.

[0085] This control law guarantees that:

[0086] 1. Fast convergence of preset formations;

[0087] 2. The speed of the cluster nodes is consistent with their heading;

[0088] 3. Maintaining formation stability in complex flight scenarios.

[0089] The formation convergence criterion is: the error of the nodes relative to the preset formation position. Preset threshold Speed ​​difference between nodes Preset threshold And the stable duration exceeds The formation is considered to have entered a stable state within seconds. Simultaneously, this step establishes a node status synchronization mechanism, ensuring that the status broadcast period of each node is perfectly matched with the consistency control period and subsequent node damage detection period. This provides a real-time data foundation for node failure detection in step three, enabling parameter linkage between steps two and three.

[0090] The consistency control in this step and the node failure detection in step three are executed synchronously with the same control cycle. Each time a consistency control calculation, neighbor state interaction and update is completed, a neighbor node failure detection is executed synchronously. The node state detection results in step three will be fed back to this step in real time, and the adjacency weights of abnormal nodes will be frozen immediately to avoid abnormal state data from affecting the overall convergence and stability of the formation.

[0091] Step 3: Node Damage Detection Mechanism

[0092] This step uses physical layer communication for detection. The dual-verification architecture for state-layer flight rationality detection is executed in the same cycle as the consistency control in step two. Each time neighbor information is received, a neighbor state detection is performed synchronously.

[0093] An anomaly warning is triggered when a node meets one of the following conditions:

[0094] 1. Communication timeout: Continuous If no status information is received from the target node within a heartbeat cycle, the heartbeat cycle matches the control cycle (default). ), The threshold can be adaptively adjusted according to the electromagnetic environment, and the pre-warning thresholds for link packet loss rate and signal-to-noise ratio can be set at the same time.

[0095] 2. Abnormal Status: The node's flight status exceeds the preset flight envelope, or the deviation from the formation reference status exceeds twice the safe formation distance. nominal airspeed And the duration of the abnormal state exceeds a preset threshold;

[0096] 3. Heartbeat signal loss: No heartbeat broadcast packet is received from the target node for two consecutive cycles. The heartbeat packet contains core information such as node ID, flight status, mission status, remaining energy, and link quality. Distinguish between heartbeat loss (physical layer disconnection) and communication timeout (no response at the data link layer).

[0097] After triggering an anomaly warning, a misjudgment avoidance mechanism is executed: the node that detects the anomaly first sends a query frame to the target node and starts the backup node pre-synchronization process of the target node, without directly determining that the node is faulty; the local voting judgment mechanism is only started when multiple neighboring nodes detect the target node as abnormal.

[0098] Neighboring nodes use a local voting mechanism to determine node failure. The voting rule is: within the two-hop neighbor range of the failed node, more than 50% of the valid nodes vote to confirm the node's failure, thus ultimately confirming the node's damage. Invalid.

[0099] Once a node failure is confirmed, neighboring nodes immediately freeze the node's adjacency weight, cease information exchange, and simultaneously broadcast the failed node ID to their 2-hop neighbor range, directly triggering the topology self-reconstruction process in step four, thus achieving a strong triggering association between steps three and four.

[0100] After the node failure is finally confirmed through the local voting mechanism in step three, the topology self-reconstruction process in step four will be triggered immediately. During the topology reconstruction process, the updated node list will be synchronously fed back to this step to update the failure detection list in real time and avoid repeated detection of nodes that have been confirmed to be failed. After the topology reconstruction is completed, the new topology parameters will be synchronized to the consistency control module in step two to restore the stable closed-loop control of the formation.

[0101] Step 4: Topology Self-Reconfiguration Mechanism

[0102] This step follows the execution priority of "first ensure connectivity, then optimize formation, and finally adjust tasks." Once a node failure is confirmed, topology self-reconstruction is immediately performed according to the following process:

[0103] 1. Update the adjacency matrix : Delete the row and column corresponding to the failed node, and add new adjacent edges to the neighboring nodes of the original failed node to ensure the number of neighbors for all valid nodes. ,satisfy The underlying requirements for connectivity;

[0104] 2. Reconstruct the Laplace matrix Based on the updated adjacency matrix Calculate the new degree matrix Generate the reconstructed Laplace matrix Immediately calculate the new algebraic connectivity. And verify If the connectivity check does not meet the requirements, continue to add adjacent edges until the connectivity check passes.

[0105] 3. Recalculate the local neighbor set The decentralized update rule is adopted, and the node only needs to interact with the information of the 1-hop neighbor to complete the neighbor set update. No global communication is required. The updated neighbor set is synchronously fed back to the consistency control module in step two to update the control law parameters in real time.

[0106] 4. Automatic recovery Connectivity Structure: After completing the above updates, perform a reconstruction convergence check. When all node degrees... , And if it remains stable for two control cycles, then the topology reconstruction is considered complete.

[0107] After the topology reconstruction is completed, the formation gap filling operation is executed synchronously. Based on the preset benchmark formation, the principle of minimum displacement adjustment is adopted. The neighboring nodes around the gap fill the formation gap with the minimum trajectory change, ensuring that the overall heading and mission direction of the cluster remain unchanged. The adjusted formation parameters are synchronously updated to the consistency control module in step two to achieve smooth convergence of the formation.

[0108] Simultaneously triggering preparatory steps for task allocation reconfiguration: the attack simulation task originally undertaken by the failed node is used to lock the nearest neighboring node with matching remaining capabilities as a candidate takeover node, providing a preparatory foundation for task redistribution in step five.

[0109] After the topology reconstruction is completed, the formation converges stably, and the connectivity verification is passed in this step, a reconstruction completion signal will be broadcast to trigger the task adaptive reallocation in step five. The result of the task reallocation will be synchronously fed back to this step. If the task execution requires adjustment of the relative positions of nodes, the local topology structure will be optimized synchronously to ensure the topology resilience throughout the entire task execution process.

[0110] Step 5: Adaptive Task Reassignment

[0111] This step uses a distributed task auction algorithm, which only redistributes unfinished tasks of failed nodes. The auction scope is limited to the two-hop neighbors of the failed node, eliminating the need for global auctions, reducing communication overhead, and adapting to decentralized architectures.

[0112] Each participating node calculates the task execution cost function, expressed as:

[0113]

[0114] in: The distance cost from the node to the task execution point. The remaining energy cost for a node to perform a task. , ,satisfy + Priority will be given to large-scale surprise attack drills. Long-endurance training missions will be given priority. .

[0115] The cost function undergoes dimensional normalization, which normalizes the distance cost. Energy cost Unified mapping to The interval is used to avoid the impact of dimensional differences on the effectiveness of weights; for multi-payload target scenarios, the cost function can be expanded to... ,in To determine the cost of mission capabilities, match mission adaptation parameters such as payload type and remaining simulated attack count.

[0116] like Figure 5 As shown, the complete execution flow of distributed task auction is as follows:

[0117] 1. Auction Initiation: The original backup node of the failed node acts as the initiator, broadcasting the failed node's unfinished task information, task deadline, and execution requirements to its 2-hop neighbors;

[0118] 2. Bidding Price: Nodes that meet the task execution requirements calculate their own cost. The bid information is broadcast to the initiator within a preset time window;

[0119] 3. Auction Conclusion: After receiving all bids, the initiating party selects the price. The smallest node acts as the task taker and broadcasts the winning bid results to the bidding area;

[0120] 4. Task handover: The winning node synchronizes all task parameters, updates its own task queue, broadcasts a task takeover confirmation signal to the entire cluster, and completes task redistribution.

[0121] After task redistribution is completed, the task parameters of the takeover node are synchronized to the consistency control module in step two, and the formation flight parameters are adjusted to match the task execution requirements. The cluster reverts to the normalized distributed consistency formation control and node status detection process of step two. When the cluster detects abnormal situations such as node failure or communication link interruption again, it will automatically repeat the entire process of "failure detection - topology reconstruction - task redistribution". Throughout the entire task execution cycle, the process proposed in step one... Connectivity structure and algebraic connectivity constraints are always the underlying rules for cluster operation, ensuring the cluster's topology resilience and task execution continuity throughout the process.

[0122] Example

[0123] In this embodiment, the following is adopted: Fixed-wing target drones were clustered together to perform simulated saturation attack training missions. The target drones' nominal airspeed was... Minimum turning radius Minimum safe flight distance The cluster communication adopts a self-organizing networking mode and operates in a complex electromagnetic interference training environment.

[0124] During the cluster initialization phase, step one, the robust topology construction, is performed, using a 3-connected graph structure, i.e. Ensure that each node is at least connected to Each neighboring node maintains a communication link and can tolerate up to [number] simultaneous [nodes]. One node failed; the initial topology adopted a diamond-shaped formation structure to complete the adjacency matrix. With Laplace matrix After construction, verify the algebraic connectivity. The connectivity requirements are met, and the topology initialization is completed.

[0125] After the topology verification passes, start the distributed consistency control in step two, with the control period set to... A cascaded consistency control architecture is adopted, with the inner speed loop ensuring consistent airspeed and heading of the cluster, and the outer position loop ensuring the relative position stability of the diamond formation. The node status broadcast period is consistent with the control period, and the node damage detection in step three is initiated synchronously. The heartbeat period is matched with the control period, and the communication timeout threshold is set to continuous. No response for one cycle.

[0126] During the exercise, within the cluster The target drone was simulated to be shot down by the air defense system, and neighboring nodes were continuously... No such period has been received. The heartbeat signals of each node were simultaneously detected, and its state was detected to be outside the flight envelope, triggering an anomaly warning; through A local voting mechanism that jumps to the neighbor range, exceeding... Valid node determination If a node fails, the failure is confirmed, and the topology self-reconstruction in step four is immediately triggered.

[0127] In the topology reconstruction phase, the adjacency matrix is ​​updated first. ,delete For each failed node, its corresponding row and column are used to add new adjacent edges to its original neighboring nodes, ensuring the remaining nodes... The number of neighbors of each node Reconstructing the Laplace matrix Verify algebraic connectivity To meet connectivity requirements, the neighbor sets of each node are updated and synchronously fed back to the consistency control module. Simultaneously, using the minimum displacement principle, neighboring nodes around the gaps complete the formation filling, smoothly adjusting the diamond-shaped formation to a trapezoidal formation, with the formation error within [a certain range]. The system converges to a preset threshold, completing the topology and formation reconstruction.

[0128] After the topology reconstruction is completed, step five, adaptive task reallocation, is triggered. Simulated penetration attack mission for each failed node, in its Initiate a distributed auction within the neighbor range, with the cost function weights set to... , Prioritize matching based on distance cost, and ultimately choose the closest match. The neighboring nodes take over the attack tasks of the failed nodes, complete the task handover and redistribution, and the cluster returns to normal formation control to continue to carry out saturation attack training tasks.

[0129] The simulation and flight test results of this embodiment show that after node failure, the convergence time of the cluster formation error is... Algebraic connectivity of the entire process The value remained consistently greater than 0, and the cluster did not experience any disintegration or instability issues, resulting in a high completion rate for the final training mission. It fully realizes the ability to withstand damage and self-reconstruction in complex adversarial environments.

Claims

1. A method for controlling a fixed-wing target drone swarm with survivable topology and self-reconfiguration capabilities, characterized in that, Includes the following steps: 1) Construct a robust topology, abstracting the fixed-wing target drone cluster into an undirected graph model, and simultaneously constructing... - A connected graph structure ensures that each node in the cluster is connected to at least one other node. Each neighboring node maintains a communication link, and the cluster connectivity is verified by the algebraic connectivity of the Laplace matrix; 2) Distributed consensus control: Local consensus protocol is used for grouping control of each target machine; 3) A dual verification architecture of physical layer communication detection and state layer flight rationality detection is adopted for node damage detection. When a node abnormality is detected, the final determination of node failure is completed through the local voting mechanism of neighboring nodes. 4) Once a node failure is confirmed, perform topology self-reconstruction, update the adjacency matrix and Laplace matrix, re-verify the algebraic connectivity, and restore the topology. - Connect the structure and synchronously fill the formation gaps using the principle of minimum displacement adjustment; 5) A distributed task auction algorithm is used for adaptive task reallocation. Only the unfinished tasks of failed nodes are reallocated within their 2-hop neighbors. The nodes participating in the auction select the task takeover node based on the cost function.

2. The fixed-wing target drone swarm control method with anti-destruction topology and self-reconfiguration capability according to claim 1, characterized in that, In the survivability topology, the nodes of the undirected graph model are individual target machines, and the edges are bidirectional communication links between nodes. The construction process is as follows: After the cluster powers on, it performs initial neighbor discovery, generating an initial adjacency matrix based on communication distance, link signal-to-noise ratio, and relative flight stability. Based on adjacency matrix Calculate the degree matrix This leads to the generation of the Laplace matrix. Based on the Laplace matrix Perform topology pre-validation to confirm whether the initial topology meets the requirements. If the connectivity requirement is not met, the adjacent edges are re-optimized until the verification is passed. The - Connected graph structure The value represents the maximum number of failed nodes the cluster can tolerate simultaneously, in relation to the cluster size. Strong binding, when the cluster size hour ,when hour ,and Meanwhile, the topology construction also satisfies the hard constraints of the fixed-wing target drone.

3. The fixed-wing target drone swarm control method with anti-destruction topology and self-reconfiguration capability according to claim 2, characterized in that, The adjacency matrix Specifically: in, Representation matrix The Line 1 Column elements, , They represent the first frame and the first Spatial coordinates of the target launcher Indicates communication distance. This represents the Euclidean distance between the two target drones. Indicates the first frame and the first Signal-to-noise ratio of the link between target drones This indicates the preset signal-to-noise ratio threshold.

4. The fixed-wing target drone swarm control method with anti-destruction topology and self-reconfiguration capability according to claim 1, characterized in that, The method of verifying cluster connectivity using the algebraic connectivity of the Laplace matrix specifically involves: Where L is the Laplacian matrix corresponding to the clustered undirected graph. It is the second smallest eigenvalue of the Laplace matrix, i.e., the algebraic connectivity.

5. A fixed-wing target drone swarm control method with survivable topology and self-reconfiguration capability according to claim 1, characterized in that, The construction of the undisturbed topology structure satisfies the following three supporting mechanisms: Multiple redundant communication paths: ensuring that at least one communication path exists between any two nodes within the cluster. Disjoint communication paths, with path priority based on link signal-to-noise ratio. Number of jumps Long-term stability ranking; automatically switches to a backup path when a single link fails. Local Neighborhood Backup Node Mechanism: A backup node is selected for each node, based on the following criteria: the node with the best link quality, the highest remaining energy, and the most stable flight status among its neighbors. 1 node The backup node synchronizes the flight status, mission parameters, and formation information of the master node in real time, and has the ability to quickly take over after the master node fails. Decentralized collaborative decision-making: Employing a local consensus mechanism, cluster decisions can be completed with only a majority vote from the neighboring nodes within two hops.

6. The fixed-wing target drone swarm control method with anti-destruction topology and self-reconfiguration capability according to claim 1, characterized in that, The control law of the local consensus protocol is: in: For the first Control inputs for each target node For the first Each target machine state vector These correspond to the node's three-dimensional spatial position, airspeed, track inclination, and heading angle, respectively. Gather for the neighbors, Adjacency matrix The corresponding adjacency weights.

7. A fixed-wing target drone swarm control method with anti-destruction topology and self-reconfiguration capability according to claim 1, characterized in that, Step 3) specifically refers to: 3.1) When a node is detected to have experienced a communication timeout, abnormal status, or lost heartbeat signal, an anomaly warning is triggered and a misjudgment avoidance mechanism is executed. That is, when an abnormal node is detected, a query frame is first sent to the target node, and the backup node pre-synchronization process of the target node is started. The node is not directly judged as failed. Only when multiple neighboring nodes detect that the target node is abnormal will a local voting judgment mechanism be started to confirm the failed or damaged node. Specifically, the local voting judgment mechanism is as follows: within the two-hop neighbor range of the failed node, more than 50% of the valid nodes vote to judge the node as failed, that is, the node is finally confirmed as damaged or failed. 3.2) After a node failure is confirmed, neighboring nodes immediately freeze the adjacency weight of the node, stop information exchange, broadcast the failed node ID to the range of 2-hop neighbors, and trigger the topology self-reconstruction process.

8. A fixed-wing target drone swarm control method with survivable topology and self-reconfiguration capability according to claim 1, characterized in that, Step 4) specifically refers to: 4.1) Update the adjacency matrix : Delete the row and column corresponding to the failed node, and add new adjacent edges to the neighboring nodes of the original failed node to ensure that the number of neighbors of all valid nodes is equal. ,satisfy The underlying requirements for connectivity; 4.2) Reconstructing the Laplace matrix Based on the updated adjacency matrix Calculate the new degree matrix Generate the reconstructed Laplace matrix Immediately calculate the new algebraic connectivity. And verify If the verification requirements are not met, continue to add adjacent edges until the connectivity verification passes. 4.3) Recalculate the local neighbor set The decentralized update rule is adopted, so that nodes only interact with information with their one-hop neighbors; 4.4) Automatic recovery Connectivity structure: when all nodes have degree , The topology reconstruction is considered complete when the topology remains stable for two consecutive control cycles.

9. A fixed-wing target drone swarm control method with survivable topology and self-reconfiguration capability according to claim 1, characterized in that, The distributed task auction algorithm is as follows: Auction Initiation: The original backup node of the failed node acts as the initiator, broadcasting the failed node's unfinished task information, task deadline, and execution requirements to its 2-hop neighbors. Bidding price: Calculate the cost of nodes that meet the task execution requirements. The bid information is broadcast to the initiator within a preset time window; Auction Decision: After receiving all bids, the initiating party selects the price. The smallest node acts as the task taker and broadcasts the winning bid results to the bidding area; Task handover: The winning node synchronizes all task parameters, updates its own task queue, broadcasts a task takeover confirmation signal to the entire cluster, and completes task redistribution.

10. A fixed-wing target drone swarm control method with survivable topology and self-reconfiguration capability according to claim 9, characterized in that, The task execution cost function is as follows: in: The distance cost from the node to the task execution point. The remaining energy cost for a node to perform a task. , The weighting coefficients for distance cost and energy cost are respectively, satisfying... + ; The cost function undergoes dimensional normalization, reducing the distance cost. Energy cost Unified mapping to For multi-payload target scenarios, the extended cost function will be... ,in, For the cost of mission capabilities, This is the weighting coefficient for the cost of task capability.

Images