A multi-unmanned aerial vehicle formation event-triggered consensus control method based on a neural tangent kernel

By introducing the event triggering mechanism of the neural tangent nucleus and adaptively adjusting the communication frequency, the problem of limited communication bandwidth in the control of multiple UAV formations is solved, and efficient, stable system operation and intelligent control are achieved.

CN121143453BActive Publication Date: 2026-06-23GUANGDONG UNIV OF TECH
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511261474.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-09-05
Publication Date
2026-06-23
Estimated Expiration
2045-09-05

AI Technical Summary

Technical Problem

Existing multi-UAV swarm control methods suffer from bandwidth constraints in industrial environments with limited communication bandwidth. Frequent communication leads to bandwidth shortages, data packet loss, and system crashes. Furthermore, traditional methods are difficult to adapt and adjust adaptively, failing to achieve an optimal balance between system performance and communication resources, and lacking intelligence and adaptability.

Method used

An event triggering mechanism based on the neural tangent kernel is adopted. The dynamic characteristics of the multi-UAV system are extracted through the neural network, and an adaptive event triggering function is designed. State information sampling and communication are only performed when the error exceeds the adaptive threshold. Combined with Bayesian optimization to adjust parameters, communication resource allocation is optimized, and it is integrated with the consistency control protocol.

Benefits of technology

It significantly reduces communication volume, improves system communication efficiency and robustness, adapts to complex industrial environments, and has good engineering feasibility and application value.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121143453B_ABST
    Figure CN121143453B_ABST
Patent Text Reader

Abstract

The application discloses a kind of multi-unmanned aerial vehicle formation event trigger consistency control methods based on neural tangent kernel, comprising the following steps: according to multi-unmanned aerial vehicle cooperative formation scene, linear system dynamics model is established to each unmanned aerial vehicle;According to the linear system dynamics model, introduce neural tangent kernel mechanism, extract the dynamic characteristics of multi-unmanned aerial vehicle;Based on the neural tangent kernel mechanism, design adaptive event trigger function, determine whether to trigger the information interaction and control update between unmanned aerial vehicles;According to the result of information interaction and control update, the parameter of adaptive event trigger mechanism is dynamically adjusted, and the allocation of communication resources between multi-unmanned aerial vehicles is optimized;According to the optimized parameter, adaptive event trigger mechanism is fused with multi-unmanned aerial vehicle consistency control protocol, and the formation control of multi-unmanned aerial vehicle is realized.The application can solve the key problems such as communication bottleneck and system robustness faced by multi-unmanned aerial vehicle formation control in actual industrial scene, and has significant technological progress.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of multi-UAV formation control technology, and particularly relates to a multi-UAV formation event triggering consistency control method based on neural tangent kernel. Background Technology

[0002] In drone swarm control scenarios, multiple drones need to maintain relative formations via wireless communication to achieve tasks such as collaborative reconnaissance and material transport. However, in real industrial environments, communication bandwidth is limited, significantly increasing the computational and communication costs of data transmission. Excessively high communication frequencies not only lead to bandwidth resource constraints but may also cause data packet loss and communication delays, severely impacting the formation maintenance and task coordination capabilities of multi-drone systems. Therefore, effectively reducing communication and computational costs while ensuring system consistency has become a critical technical challenge that urgently needs to be addressed in the field of multi-drone swarm control.

[0003] Traditional multi-drone swarm control methods, such as consensus algorithms, time-triggered methods, continuous event-triggered methods, self-triggered methods, and dynamic event-triggered methods, typically assume that drones can communicate frequently or even continuously. However, in real-world industrial scenarios, communication bandwidth is limited, and frequent communication can lead to bandwidth resource constraints, easily causing communication congestion, data packet loss, and even system crashes.

[0004] While existing event-triggered mechanisms can reduce communication to some extent, most rely on fixed thresholds or simple triggering conditions, making it difficult to adaptively adjust to changes in system state, environment, and task requirements. In complex or dynamically changing industrial environments, fixed event-triggered mechanisms may result in excessively high or low communication frequencies, failing to achieve an optimal balance between system performance and communication resources.

[0005] Furthermore, traditional control methods are mostly based on linear system assumptions, making it difficult to address the nonlinear, time-varying, and complex coupling problems commonly found in real-world multi-UAV systems. Existing solutions primarily rely on manually designed control laws and triggering conditions, making it difficult to fully utilize actual operational data for self-optimization and learning, and lacking intelligence and adaptive capabilities.

[0006] To address the aforementioned problems in existing technologies, there is an urgent need to propose a multi-UAV formation event-triggered consistency control method based on neural tangent kernels. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a multi-UAV formation event-triggered consistency control method based on neural tangent kernels. This method combines event-triggered mechanisms with the efficient feature representation capabilities of neural networks. By designing adaptive event-triggered conditions, each UAV only samples and communicates its state information when the error exceeds an adaptive threshold, thereby significantly reducing unnecessary data transmission and computational burden. This solution not only improves the system's communication efficiency and reliability but also possesses good engineering feasibility and application value, making it widely applicable to UAV formations.

[0008] This invention proposes a multi-UAV formation event-triggered consistency control method based on neural tangential kernels, comprising the following steps:

[0009] Based on the scenario of multi-UAV cooperative formation, a linear system dynamics model is established for each UAV;

[0010] Based on the linear system dynamics model, a neural tangent kernel mechanism is introduced to extract the dynamic characteristics of multiple unmanned aerial vehicles (UAVs).

[0011] Based on the aforementioned neural tangent kernel mechanism, an adaptive event triggering function is designed to determine whether to trigger information interaction and control updates between UAVs.

[0012] Based on the results of information interaction and control updates, the parameters of the adaptive event triggering mechanism are dynamically adjusted to optimize the allocation of communication resources among multiple UAVs.

[0013] Based on the optimized parameters, the adaptive event triggering mechanism is integrated with the multi-UAV consensus control protocol to achieve formation control of multiple UAVs.

[0014] Optionally, based on the linear system dynamics model, the process of introducing a neural tangent kernel mechanism to extract the dynamic characteristics of multiple unmanned aerial vehicles includes:

[0015] By performing eigenvalue decomposition on the neural tangent nucleus, extracting the principal eigenvalues ​​and corresponding eigenvectors, the nonlinear characteristics and state change trends of multi-UAV systems can be identified.

[0016] Optionally, the formula for the neural tangential nucleus mechanism is as follows:

[0017]

[0018] Among them, f i (x i (t),θ i Let θ be the neural network output of the i-th drone. i These are network parameters.

[0019] Optionally, based on the neural tangent kernel mechanism, an adaptive event triggering function is designed to determine whether to trigger information interaction and control updates between UAVs, including:

[0020] When the change in the neural tangent nucleus exceeds the dynamic adaptive threshold, it triggers information interaction and control updates between drones.

[0021] Optionally, the formula for the adaptive event triggering function is as follows:

[0022] T i (t)=||K i (t)-K i (t last,i )||-η i (t)>0,

[0023] Among them, t last,i Let η be the time when the i-th drone last triggered communication. i (t) is the dynamic adaptive threshold function.

[0024] Optionally, a dynamic adaptive threshold function is constructed based on the functional relationship between the initial threshold, the sensitivity coefficient of the tangential nucleus, and the time decay coefficient; and the initial threshold, the sensitivity coefficient of the tangential nucleus, and the time decay coefficient are adaptively adjusted based on the introduction of a Bayesian optimization algorithm based on the historical change pattern of the tangential nucleus.

[0025] Optionally, the control strategy that integrates the adaptive event triggering mechanism with the multi-UAV consensus control protocol is shown in the following equation:

[0026] The event triggering indicator function for the tangential nucleus is:

[0027]

[0028] The distributed consensus control input for the i-th drone is:

[0029] s(t)))+(1-δ i (t))u i (t last,i ),

[0030] Where, δ i (t) Event triggering function of the neural tangential nucleus, N i Let i be the set of communication neighbors of the i-th drone. Let ζ(k) be the neighbor weights, and ζ(k) be the topology switching probability function. The position of drone i at the trigger time. h represents the position of drone j at the trigger time. i For the desired formation position, hj For the formation positions of adjacent nodes, g i Let be the coupling gain between the i-th UAV and the reference signal s(t), where s(t) is the global reference signal. U represents the most recent event trigger time for the i-th drone. i (t last,i ) indicates the previous control output.

[0031] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0032] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method.

[0033] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method.

[0034] Compared with the prior art, the present invention has the following advantages and technical effects:

[0035] (1) Communication efficiency has been greatly improved, meeting the actual industrial needs:

[0036] Existing technologies often employ fixed thresholds or simple event-triggered mechanisms, which struggle to adapt to dynamic system changes, leading to excessively high or low communication frequencies and failing to balance system performance and bandwidth resources. This invention, through an NTK-driven adaptive event-triggered mechanism, achieves intelligent adjustment of the communication frequency, triggering communication only when system dynamics change significantly. This drastically reduces communication volume and greatly alleviates bandwidth pressure and data congestion risks in real-world industrial scenarios.

[0037] (2) The ability to model and control complex systems is significantly enhanced:

[0038] Traditional methods are mostly based on linear system assumptions, making it difficult to address the nonlinear, time-varying, and complex coupling problems commonly found in real-world multi-UAV systems. This invention introduces the neural tangent kernel theory and utilizes neural networks to perform high-dimensional modeling of the system's complex dynamics. This enables the accurate capture of the system's nonlinear characteristics and dynamic trends, significantly improving the system's modeling and control capabilities.

[0039] (3) High level of self-adaptability and intelligence, and strong robustness:

[0040] This invention achieves intelligent optimization of the event-triggered mechanism through machine learning methods such as adaptive threshold dynamic adjustment and Bayesian optimization. It can dynamically adjust parameters based on historical data and actual operating status, improving the system's adaptability and robustness. Even when faced with environmental disturbances, model uncertainties, or task changes, the system can still maintain efficient and stable operation.

[0041] (4) High feasibility of engineering implementation, and parameters can be flexibly configured:

[0042] This invention fully considers engineering parameters such as bandwidth, latency, and error thresholds in real-world industrial scenarios. All key parameters can be flexibly set according to actual needs, demonstrating excellent engineering feasibility and application value. In contrast, existing technologies often lack detailed consideration of actual industrial parameters, resulting in limited practicality.

[0043] (5) High system integration:

[0044] This invention not only proposes an NTK-driven event triggering mechanism in theory, but also provides a complete system implementation architecture, covering the entire process of state acquisition, feature extraction, event triggering, parameter optimization, and control protocol, which facilitates practical engineering implementation and large-scale application. Attached Figure Description

[0045] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0046] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention. Detailed Implementation

[0047] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0048] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0049] Example 1

[0050] like Figure 1 As shown, this embodiment provides a multi-UAV formation event triggering consistency control method based on neural tangent kernel, including the following steps:

[0051] Based on the scenario of multi-UAV cooperative formation, a linear system dynamics model is established for each UAV;

[0052] Based on the linear system dynamics model, a neural tangent kernel mechanism is introduced to extract the dynamic characteristics of multiple unmanned aerial vehicles (UAVs).

[0053] Based on the aforementioned neural tangent kernel mechanism, an adaptive event triggering function is designed to determine whether to trigger information interaction and control updates between UAVs.

[0054] Based on the results of information interaction and control updates, the parameters of the adaptive event triggering mechanism are dynamically adjusted to optimize the allocation of communication resources among multiple UAVs.

[0055] Based on the optimized parameters, the adaptive event triggering mechanism is integrated with the multi-UAV consensus control protocol to achieve formation control of multiple UAVs.

[0056] The feasible process for linear system modeling of multi-UAV systems includes:

[0057] For scenarios involving multiple drones in coordinated formation, the following linear system dynamics model is first established for each drone:

[0058]

[0059] in, Let represent the state variables (such as position, velocity, attitude, etc.) of the i-th UAV at time t. To control inputs (such as motor thrust, control surface deflection angle, etc.), A i B is the system dynamics matrix. i The model serves as the control input matrix. It accurately describes the state evolution of multiple UAVs in industrial scenarios such as formation flying and formation maintenance, providing a theoretical foundation for subsequent intelligent control and event triggering mechanism design.

[0060] This embodiment establishes a general linear system dynamics model for multi-drone collaborative scenarios involving drones and robots, clarifying the state variables, control inputs, and their physical meanings for each drone, thus providing a theoretical foundation for subsequent intelligent control and event triggering mechanisms. Through a discrete sampling mechanism, it can adapt to the objective limitations of sensor sampling intervals and communication discontinuities in real-world industrial scenarios.

[0061] The feasible process of online computation and system feature extraction of the Neural Tangent Kernel (NTK) includes:

[0062] To enhance the system's adaptability to complex and nonlinear dynamics, this embodiment introduces a neural tangent kernel (NTK) mechanism on each UAV. Let f i (xi (t),θ i Let θ be the neural network output of the i-th drone. i For network parameters, NTK is defined as follows:

[0063]

[0064] This kernel function characterizes the sensitivity of the neural network's parameters to changes in the output, dynamically reflecting the system's nonlinear characteristics and state change trends. Through K... i (t) Perform eigenvalue decomposition to extract the principal eigenvalue λ. i,j (t) and the corresponding eigenvector V i,j (t) enables automatic identification of the main dynamic characteristics of the system, providing data support for the subsequent adaptive event triggering mechanism.

[0065] This invention introduces online computation of the neural tangent kernel on each UAV, utilizing neural networks to model the high-dimensional dynamic features of complex systems. By decomposing the eigenvalues ​​of the NTK matrix, the main dynamic characteristics of the system are extracted in real time, achieving accurate characterization of the nonlinear and time-varying characteristics of multi-UAV systems and greatly improving the system's adaptability to complex environments.

[0066] The feasible process for an adaptive event triggering mechanism based on NTK changes includes:

[0067] To significantly reduce communication overhead and improve system real-time performance and engineering applicability, this embodiment designs an adaptive event triggering function based on NTK changes:

[0068] T i (t)=||K i (t)-K i (t last,i )||-η i (t)>0,

[0069] Among them, t last,i Let η be the time when the i-th drone last triggered communication. i (t) is the dynamic adaptive threshold function. When the change in NTK exceeds the threshold, i.e., T... i If (t)>0, information exchange and control updates between UAVs are triggered. This mechanism can sensitively and adaptively adjust the communication frequency according to dynamic changes in the system, effectively avoiding unnecessary high-frequency data transmission and greatly alleviating bandwidth pressure.

[0070] This embodiment designs an adaptive event triggering function based on NTK changes, which can intelligently determine when to trigger a communication event according to the actual needs of dynamic system changes. This mechanism not only considers the magnitude of NTK changes, but also introduces an adaptive threshold dynamic adjustment strategy, making communication triggering more scientific and reasonable, and significantly reducing communication frequency and bandwidth consumption.

[0071] An feasible process for adaptive dynamic adjustment of the threshold function includes:

[0072] To further enhance the intelligence and robustness of the event triggering mechanism, this embodiment proposes the following adaptive threshold design:

[0073]

[0074] Where, η 0,i Let α be the initial threshold. i β is the NTK sensitivity coefficient. i The time decay coefficient, The Frobenius norm derivative of NTK reflects the dynamic rate of change of the system. This design allows the threshold to automatically decrease when the UAV system state changes rapidly, improving communication and control response speed; when the system tends to stabilize, the threshold automatically increases, reducing unnecessary communication and further optimizing bandwidth utilization.

[0075] An implementable, prediction-based process for optimizing the allocation of communication resources includes:

[0076] Considering the limited and unpredictable communication resources in industrial scenarios, this embodiment introduces a Bayesian optimization algorithm based on historical NTK variation patterns, adjusting the parameter η. 0,i α i β i The system will make adaptive adjustments. By analyzing and predicting historical data, the parameters of the event triggering mechanism will be dynamically optimized to achieve optimal allocation of communication resources and ensure that the system can operate efficiently and stably under different operating conditions.

[0077] This embodiment proposes an adaptive threshold dynamic adjustment algorithm based on NTK derivatives, combined with machine learning methods such as Bayesian optimization, to dynamically optimize event triggering parameters based on historical data, achieving optimal allocation of communication resources. This mechanism can flexibly adjust parameters such as bandwidth and latency according to actual industrial scenarios, ensuring efficient and stable system operation.

[0078] The feasible process of integrating an adaptive event-triggered mechanism with a multi-UAV consensus control protocol to achieve multi-UAV formation control, based on optimized parameters, includes:

[0079] This embodiment deeply integrates the NTK event triggering mechanism with the multi-UAV consensus control protocol, and proposes the following novel control strategy:

[0080] First, define the NTK event trigger indicator function:

[0081]

[0082] The distributed consensus control input for the i-th drone is:

[0083]

[0084] Where, N i Let i be the set of communication neighbors of the i-th drone. h represents the neighbor weights. i Let s(t) be the desired formation position, and s(t) be the global reference signal. Indicates the most recent event trigger time for the i-th drone; u i (t last,i The last control output is indicated by the symbol ). This protocol ensures that the UAV only performs information acquisition, communication, and control updates when the NTK event triggering conditions are met; otherwise, it retains the last control command, thus significantly reducing the system's communication burden. This fusion strategy fully considers both the system's dynamic response capabilities and practical engineering communication constraints, exhibiting good robustness and practicality.

[0085] Upon event triggering, this embodiment employs a distributed consensus control protocol to achieve high-precision formation maintenance and collaborative control. This protocol fully considers parameters such as communication topology, neighbor weights, and desired formation in real-world industrial scenarios, demonstrating good engineering feasibility and widespread application value.

[0086] As a specific implementation method, this embodiment combines a Neural Tangent Kernel (NTK) with an event-triggered mechanism to design more intelligent and adaptive triggering conditions. This ensures that the UAV communicates only when necessary, greatly reducing unnecessary data transmission. This helps ensure efficient collaboration and stable operation of multi-UAV systems in real-world industrial environments with limited bandwidth and discontinuous communication. The NTK possesses powerful nonlinear modeling capabilities, enabling it to better capture complex dynamic relationships and high-dimensional state characteristics in multi-UAV systems. Combined with the event-triggered mechanism, it can achieve efficient control of complex systems such as nonlinear, time-varying, and multi-coupled systems, overcoming the limitations of traditional linear methods. Utilizing the learning capabilities of neural networks, the event-triggered conditions can be dynamically optimized based on actual operating data, achieving adaptive adjustment and balancing system performance with optimal communication resource consumption. This allows the system to intelligently adjust communication and control strategies based on task requirements, environmental changes, and other factors, improving the system's robustness and adaptability.

[0087] Considering the overall system architecture and application scenarios, this embodiment not only proposes a theoretical model, but also provides a complete system implementation architecture, covering the entire process of state acquisition, NTK calculation, event triggering, parameter optimization, control protocol, etc., and is applicable to a variety of practical industrial scenarios such as UAV formation, intelligent vehicle queue, and spacecraft collaboration.

[0088] Example 2

[0089] This embodiment also discloses a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in Embodiment 1.

[0090] Example 3

[0091] This embodiment also discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in Embodiment 1.

[0092] Example 4

[0093] This embodiment also discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in Embodiment 1.

[0094] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for event-triggered consistency control of multi-UAV formations based on neural tangent kernels, characterized in that, Includes the following steps: Based on the scenario of multi-UAV cooperative formation, a linear system dynamics model is established for each UAV; Based on the linear system dynamics model, a neural tangent kernel mechanism is introduced to extract the dynamic characteristics of multiple unmanned aerial vehicles (UAVs). Based on the aforementioned neural tangent kernel mechanism, an adaptive event triggering function is designed to determine whether to trigger information interaction and control updates between UAVs. Based on the results of information interaction and control updates, the parameters of the adaptive event triggering mechanism are dynamically adjusted to optimize the allocation of communication resources among multiple UAVs. Based on the optimized parameters, the adaptive event triggering mechanism is integrated with the multi-UAV consensus control protocol to achieve formation control of multiple UAVs; Based on the aforementioned linear system dynamics model, the process of extracting the dynamic characteristics of multiple unmanned aerial vehicles (UAVs) by introducing the neural tangent kernel mechanism includes: By performing eigenvalue decomposition on the neural tangent nucleus, extracting the principal eigenvalues ​​and corresponding eigenvectors, the nonlinear characteristics and state change trends of multi-UAV systems can be identified. Based on the aforementioned neural tangent kernel mechanism, an adaptive event triggering function is designed. The process of determining whether to trigger information interaction and control updates between UAVs includes: When the change in the neural tangent nucleus exceeds the dynamic adaptive threshold, it triggers information interaction and control updates between drones. The formula for the adaptive event triggering function is shown below: , in, Let i be the time when the i-th drone last triggered communication. It is a dynamic adaptive threshold function; Based on the functional relationship between the initial threshold, the sensitivity coefficient of the tangential nucleus, and the time decay coefficient, a dynamic adaptive threshold function is constructed; and based on the introduction of a Bayesian optimization algorithm based on the historical change pattern of the tangential nucleus, the initial threshold, the sensitivity coefficient of the tangential nucleus, and the time decay coefficient are adaptively adjusted. The formula for the neural tangential nucleus mechanism is as follows: , in, This is the neural network output for the i-th drone. These are network parameters.

2. The method according to claim 1, characterized in that, The control strategy, which integrates the adaptive event triggering mechanism with the multi-UAV consensus control protocol, is shown in the following equation: The event triggering indicator function for the tangential nucleus is: , The distributed consensus control input for the i-th drone is: , in, Event trigger function of the tangential nucleus. Let i be the set of communication neighbors of the i-th drone. For neighbor weights, To switch the topological probability function, The position of drone i at the trigger time. The position of drone j at the trigger time. For the desired formation position, The formation positions of adjacent nodes. For the i-th UAV and the reference signal Coupling gain, As a global reference signal, This represents the time when the most recent event was triggered for the i-th drone. This indicates the previous control output.

3. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-2.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1-2.

5. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1-2.

Citation Information

Patent Citations

  • Multi-rotor unmanned aerial vehicle time-varying formation control method and system with input saturation

    CN113342037A

  • Spacecraft cluster robust formation control method based on dynamic event triggering mechanism

    CN120428755A