A distributed data-driven model-free adaptive output tracking control method and system for a UAV swarm

Abstract

This invention discloses a distributed data-driven model-free adaptive output tracking control method and system for UAV swarms, belonging to the field of UAV swarm cooperative control. The main steps are: 1. Constructing a dynamic linearized model using pseudo-biased steering variables. 2. Designing a batch update mechanism based on the slow time-varying assumption of pseudo-biased steering variables to reduce computational burden. 3. Designing a neighborhood output tracking error that integrates local tracking error and neighbor output consistency to construct a distributed control law. 4. Based on a directed communication topology graph, designing a distributed control law that achieves accurate tracking of time-varying reference signals by all followers, provided that only a spanning tree with the leader UAV as the root node is required. This method only requires input and output data such as the speed and altitude of the UAV and its neighbors, without requiring mathematical models or global topology information. It has relaxed requirements for communication conditions, strong applicability, and excellent tracking performance.

Description

Technical Field

[0001] This invention relates to unmanned aerial vehicle (UAV) swarm cooperative control technology, and in particular to a distributed data-driven model-free adaptive output tracking control system and method for UAV swarms. Background Technology

[0002] In recent years, unmanned aerial vehicle (UAV) swarms have been widely used in collaborative reconnaissance, search and rescue, environmental monitoring, formation flight, and target encirclement. Distributed control, as the core control paradigm of UAV swarms, requires each UAV to independently perform calculations and control decisions, relying only on its own state, measurement information, and limited information exchanged with its directly connected neighboring UAVs, in order to achieve the global objectives of the entire UAV swarm through collaborative cooperation.

[0003] Consistency is a key issue in UAV swarm control, aiming to design control strategies that rely solely on local information, enabling the swarm to reach consensus on certain quantities of interest. However, achieving accurate consistency tracking becomes particularly challenging when the follower UAV's dynamics model exhibits nonlinearity, strong coupling, and unknowns. Traditional methods typically rely on accurate aerodynamic models, but models built for real-world UAV systems often possess high-order, strongly nonlinear, and time-varying characteristics, making them difficult to use for controller design.

[0004] Data-driven control methods abandon the traditional approach of first establishing a physical model. By directly analyzing and utilizing flight data, they avoid performance bottlenecks caused by inaccurate models. Among them, model-free adaptive control, as a typical online data-driven control method, does not rely on the precise mathematical model of the UAV. Instead, it constructs an equivalent dynamic linearized model solely through real-time input / output data, thereby achieving effective control of unknown nonlinear UAV systems.

[0005] Existing research on model-free adaptive control methods for UAV swarms typically requires a strongly connected communication topology graph, and each UAV needs to know the topological information of the entire graph, which does not meet the requirements of fully distributed control. Furthermore, the pseudo-partial derivative parameter update mechanism in traditional model-free adaptive control updates parameters at every sampling time, resulting in a significant computational burden and limitations in physical systems where UAV flight parameters change slowly.

[0006] Therefore, there is an urgent need to develop a multi-UAV cooperative control method that can adapt to complex flight environments, does not rely on precise models, and possesses strong adaptive capabilities. This invention addresses this problem by proposing a multi-UAV cooperative control scheme based on model-free adaptive control. It aims to achieve high-precision and robust cooperative operations through a data-driven approach, providing technical support for the development of modern UAVs. Summary of the Invention

[0007] To address the above technical problems, this invention provides a distributed data-driven model-free adaptive output tracking control system and method for unmanned aerial vehicle (UAV) swarms.

[0008] The technical solution adopted by this invention to solve its technical problem is:

[0009] A distributed data-driven model-free adaptive output tracking control method for unmanned aerial vehicle (UAV) swarms, the method comprising the following steps:

[0010] S1: Utilize sensors to acquire the speed and altitude of each follower drone and the speed and altitude of neighboring drones in real time;

[0011] S2: Based on the acquired output data and control input data, construct a dynamic linearized data model and use pseudo-biased variables to describe the relationship between system input and output;

[0012] S3: Based on the assumption that the pseudo-biased steering variable is slowly time-varying, a batch update mechanism is designed to estimate and update the pseudo-biased steering variable every N sampling periods.

[0013] S4: Design the neighboring output tracking error, integrate the local tracking error with the consistency error of the neighboring output, and construct a distributed control law;

[0014] S5: Based on the directed communication topology graph, a distributed control law is designed to enable all follower UAVs to accurately track the output of time-varying reference signals, provided that only the spanning tree with the leader as the root node is included.

[0015] Furthermore, step S2 includes the following steps:

[0016] S2.1: For the i-th follower agent, its nonlinear discrete-time system model is expressed as:

[0017] ;

[0018] in, It is an unknown nonlinear function. and These represent the velocity command input and position output at time k, respectively. , The order is unknown;

[0019] S2.2: Under the generalized Lipschitz condition, the original nonlinear system is transformed into the following partial scheme dynamic linearized data model:

[0020] ;

[0021] in, This represents the change in velocity. Let be the pseudo-gradient parameter vector, where , This represents the change in position, where L is the pseudo-order.

[0022] Furthermore, S3 includes the following steps: Based on the physical assumption that the UAV flight parameters are slowly time-varying, a batch update mechanism is adopted, that is, the pseudo-biased directional parameters are estimated and updated every N sampling periods. Equations are constructed by collecting input and output data from multiple time points, and solved using pseudo-inverse or singular value decomposition, thereby reducing the computational burden and improving the parameter estimation accuracy. The specific process is as follows:

[0023] S3.1 Based on the physical characteristic that the flight parameters of the UAV change more slowly than the input and output, a pseudo-biased variable is assumed. The input and output variables are slowly time-varying, meaning there exists a positive integer N > 1 such that:

[0024] ;

[0025] Based on this assumption, we can obtain:

[0026] ;

[0027] The above formula is expressed over a time interval as follows:

[0028] ;

[0029] Repeat the same operation and collect the resulting I / O variables, where k > L + N:

[0030] ;

[0031] The compact form can be presented as:

[0032]

[0033] The explicit expression for an unknown PPD vector can be represented as:

[0034]

[0035] in express The pseudo-inverse matrix. When the matrix has full row rank or full column rank, its expression is: Otherwise, it can be obtained using standard techniques based on singular value decomposition.

[0036] Furthermore, step S4 includes the following steps:

[0037] S4.1 Define a directed graph , where the set of nodes Indicates the drone's serial number. Represents the set of edges in a topological graph. For a topological graph with non-negative elements The adjacency matrix represents the nodes. and Connection weights between them;

[0038] S4.2 defines the Laplace matrix of a directed graph as follows: , ,in , This is called the in-degree of vertex i. When hour, ;when hour, ;

[0039] S4.3 Define neighborhood output tracking error for:

[0040] ;

[0041] in, As a time-varying reference trajectory (desired position), when UAV i can directly access the leader UAV, and ;otherwise and These are elements of the adjacency matrix;

[0042] Furthermore, step S5 includes the following steps:

[0043] S5.1 Design Expected Trajectory: The distributed control law is:

[0044] ;

[0045] in, It is an adjustable parameter. This is the first component of the pseudo-biased variable. Under the batch update mechanism, the pre-set value is used first. The control quantity is calculated using other pseudo-partial derivative parameters, and then updated synchronously every N cycles using multi-time data. All parameters other than those mentioned above.

[0046] The present invention provides a distributed data-driven unmanned aerial vehicle swarm model-free adaptive output tracking control system, comprising a data acquisition module, a neighborhood error construction module, a batch parameter update module, a model-free adaptive control module, and a communication update module.

[0047] The data acquisition module is used to acquire the location information of each follower drone and neighboring drones in real time using the sensors on the drone.

[0048] The neighborhood error construction module is used to design a neighborhood output tracking error that integrates the local location tracking error and the location consistency error of neighboring UAVs.

[0049] The batch parameter update module is used to perform batch estimation and update of pseudo-biased parameters every N sampling periods based on the assumption of slow time-varying UAV flight parameters.

[0050] The model-free adaptive control module constructs a dynamic linearized data model based on position information and velocity command input data. It utilizes neighborhood output tracking error and pseudo-biased steering input to construct a distributed control law, calculates the velocity command input of each UAV, and achieves accurate position tracking of time-varying reference trajectory, thus realizing model-free adaptive control.

[0051] The communication update module is used to send the position and speed information of the navigator drone to the follower drone.

[0052] The beneficial effects of this invention include: The distributed data-driven UAV swarm output tracking control method proposed in this invention, according to step S2, completely eliminates the dependence on UAV aerodynamic models in controller design, requiring only the speed command inputs and position output data of each UAV and its neighbors, without any global model or topology information; according to step S3, this invention adopts a pseudo-biased variable batch update mechanism, based on the reasonable physical assumption that UAV flight parameters change slower than position changes, updating parameters every N sampling periods, significantly reducing the computational burden of the onboard computing unit, while improving parameter estimation accuracy through data accumulation; according to step S5, this invention only requires a spanning tree with the leader UAV as the root node for UAV swarm communication topology, which is more lenient than the strongly connected graph requirement of existing technologies, maintaining control performance even in the event of partial communication link failure, making it more practical for engineering applications. Attached Figure Description

[0053] Figure 1 The flowchart is a distributed data-driven model-free adaptive output tracking control method for UAV swarms according to the present invention.

[0054] Figure 2 This is a communication topology diagram of a drone swarm.

[0055] Figure 3 This is a tracking effect diagram of a drone swarm;

[0056] Figure 4 A tracking error diagram of the drone swarm;

[0057] Figure 5 This is a tracking effect diagram for the traditional model-free adaptive method. Detailed Implementation

[0058] The following refers to the accompanying drawings in the instruction manual. Figures 1-5 The technical solutions in the embodiments of the present invention will be further described in detail below.

[0059] like Figure 1 As shown, a distributed data-driven model-free adaptive output tracking control system and method for UAV swarms is disclosed. The specific implementation process of the method is as follows:

[0060] S1: Utilize sensors to acquire the speed and altitude of each follower drone and the speed and altitude of neighboring drones in real time;

[0061] S2: Based on the acquired output data and control input data, construct a dynamic linearized data model and use pseudo-biased variables to describe the relationship between system input and output;

[0062] S3: Based on the assumption that the pseudo-biased steering variable is slowly time-varying, a batch update mechanism is designed to estimate and update the pseudo-biased steering variable every N sampling periods.

[0063] S4: Design the neighboring output tracking error, integrate the local tracking error with the consistency error of the neighboring output, and construct a distributed control law;

[0064] S5: Based on the directed communication topology graph, a distributed control law is designed to enable all follower UAVs to accurately track the output of time-varying reference signals, provided that only the spanning tree with the leader as the root node is included.

[0065] Furthermore, step S2 includes the following steps:

[0066] S2.1: For the i-th follower agent, its nonlinear discrete-time system model is expressed as:

[0067] ;

[0068] in, It is an unknown nonlinear function. and These represent the velocity command input and position output at time k, respectively. , The order is unknown;

[0069] S2.2: Under the generalized Lipschitz condition, the original nonlinear system is transformed into the following partial scheme dynamic linearized data model:

[0070] ;

[0071] in, This represents the change in velocity. Let be the pseudo-gradient parameter vector, where , This represents the change in position, where L is the pseudo-order.

[0072] Furthermore, step S3 includes the following steps:

[0073] S3.1 Based on the physical characteristic that the flight parameters of the UAV change more slowly than the input and output, a pseudo-biased variable is assumed. The input and output variables are slowly time-varying, meaning there exists a positive integer N > 1 such that:

[0074] ;

[0075] Based on this assumption, we can obtain:

[0076] ;

[0077] The above formula is expressed over a time interval as follows:

[0078] ;

[0079] Repeat the same operation and collect the resulting I / O variables, where k > L + N:

[0080] ;

[0081] The compact form can be presented as:

[0082] ;

[0083] The explicit expression for an unknown PPD vector can be represented as:

[0084] ;

[0085] in express The pseudo-inverse matrix. When the matrix has full row rank or full column rank, its expression is: Otherwise, it can be obtained using standard techniques based on singular value decomposition.

[0086] Furthermore, step S4 includes the following steps:

[0087] S4.1 Define a directed graph , where the set of nodes Indicates the drone's serial number. Represents the set of edges in a topological graph. For a topological graph with non-negative elements The adjacency matrix represents the nodes. and Connection weights between them;

[0088] S4.2 defines the Laplace matrix of a directed graph as follows: , ,in , This is called the in-degree of vertex i. When hour, ;when hour, ;

[0089] S4.3 Define neighborhood output tracking error for:

[0090] ;

[0091] in, As a time-varying reference trajectory (desired position), when UAV i can directly access the leader UAV, and ;otherwise and For elements of the adjacency matrix; in Figure 2 In this communication topology, drone 0 is the leader, and drones 1, 2, 3, and 4 are followers. Drone 1 receives the location information of drone 0, drone 2 receives the location information of drone 1, drone 3 receives the location information of drone 0 and drone 2, and drone 4 receives the location information of drone 1 and drone 3.

[0092] Furthermore, step S5 includes the following steps:

[0093] S5.1 Design Expected Trajectory: ;

[0094] The distributed control law is:

[0095] ;

[0096] in, It is an adjustable parameter. This is the first component of the pseudo-biased variable. Under the batch update mechanism, the pre-set value is used first. The control quantity is calculated using other pseudo-partial derivative parameters, and then updated synchronously every N cycles using multi-time data. All parameters other than those mentioned above.

[0097] To verify the effectiveness of this method, simulations were performed using MATLAB to compare it with traditional model-free adaptive control methods. The simulation results are shown in the figure below. Figure 3 , 4 As shown in Figure 5. Figure 3 The performance of the proposed method is demonstrated. Even with time-varying desired trajectories, the follower drone can still accurately track the leader drone. Figure 4The tracking errors of the four corresponding follower drones are shown. As the control process progresses, the tracking errors of all drones stabilize within a very small range close to zero and remain at this level until the simulation ends. Figure 3 and 4 The results directly demonstrate that the proposed method has good fast convergence capability, high steady-state tracking accuracy, and effective adaptability to time-varying reference signals. Figure 5 The output performance of a UAV using a traditional model-free adaptive control method is demonstrated under the same reference trajectory. Figure 3 and Figure 5 The comparison intuitively verifies that, under the same reference trajectory conditions, this method has a smoother tracking process and higher steady-state tracking accuracy compared to the traditional model-free adaptive control method.

[0098] The present invention provides a distributed data-driven unmanned aerial vehicle swarm model-free adaptive output tracking control system, comprising a data acquisition module, a neighborhood error construction module, a batch parameter update module, a model-free adaptive control module, and a communication update module.

[0099] The data acquisition module is used to acquire the location information of each follower drone and neighboring drones in real time using the sensors on the drone.

[0100] The neighborhood error construction module is used to design a neighborhood output tracking error that integrates the local location tracking error and the location consistency error of neighboring UAVs.

[0101] The batch parameter update module is used to perform batch estimation and update of pseudo-biased parameters every N sampling periods based on the assumption of slow time-varying UAV flight parameters.

[0102] The model-free adaptive control module constructs a dynamic linearized data model based on position information and velocity command input data. It utilizes neighborhood output tracking error and pseudo-biased steering input to construct a distributed control law, calculates the velocity command input of each UAV, and achieves accurate position tracking of time-varying reference trajectory, thus realizing model-free adaptive control.

[0103] The communication update module is used to send the position and speed information of the navigator drone to the follower drone.

[0104] The modules in the aforementioned UAV control system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0105] The above provides a detailed description of the distributed data-driven model-free adaptive output tracking control system and method for UAV swarms provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention; these examples are merely illustrative to help understand the core ideas of the invention. Furthermore, this invention provides a reference method for research in the same field and can be further extended to other related UAV path tracking control fields, possessing high practicality and promotional value.

[0106] It should be noted that those skilled in the art can make various improvements and modifications to this invention without departing from its principles, and these improvements and modifications also fall within the scope of protection of the claims of this invention.

Claims

1. A distributed data-driven model-free adaptive output tracking control method for unmanned aerial vehicle (UAV) swarms, the method comprising the following steps: S1: Utilize sensors to acquire the speed and altitude of each follower drone and the speed and altitude of neighboring drones in real time; S2: Based on the acquired output data and control input data, construct a dynamic linearized data model and use pseudo-biased variables to describe the relationship between system input and output; S3: Based on the assumption that the pseudo-biased steering variable is slowly time-varying, a batch update mechanism is designed to estimate and update the pseudo-biased steering variable every N sampling periods. S4: Design the neighboring output tracking error, integrate the local tracking error with the consistency error of the neighboring output, and construct a distributed control law; S5: Based on the directed communication topology graph, a distributed control law is designed to enable all follower UAVs to accurately track the output of time-varying reference signals, provided that only the spanning tree with the leader as the root node is included.

2. The method of claim 1, wherein, S2 includes the following steps: S2.1: For the i-th follower agent, its nonlinear discrete-time system model is expressed as: ； in, It is an unknown nonlinear function. and These represent the velocity command input and position output at time k, respectively. , The order is unknown; S2.2: Under the generalized Lipschitz condition, the original nonlinear system is transformed into the following partial scheme dynamic linearized data model: ； in, This represents the change in velocity. Let be the pseudo-gradient parameter vector, where , This represents the change in position, where L is the pseudo-order.

3. The method according to claim 1, characterized in that, S3 includes the following steps: Under the assumption of slow time-varying, a batch update mechanism is adopted, and pseudo-biased directional variables are solved by using multi-time data every N periods through pseudo-inverse or singular value decomposition, so as to reduce the computational burden and improve the estimation accuracy.

4. The method according to claim 1, characterized in that, S4 includes the following steps: S4.1 Define a directed graph , where the set of nodes Indicates the drone's serial number. Represents the set of edges in a topological graph. For a topological graph with non-negative elements The adjacency matrix represents the nodes. and Connection weights between them; S4.2 defines the Laplace matrix of a directed graph as follows: , ,in , The in-degree of vertex i is called the in-degree when hour, ;when hour, ; S4.3 Define neighborhood output tracking error for: ； in, As a time-varying reference trajectory (desired position), when UAV i can directly access the leader UAV, and ;otherwise and These are the elements of the adjacency matrix.

5. The method according to claim 1, characterized in that, S5 includes the following steps: S5.1 Design Expected Trajectory: The distributed control law is: ； in, It is an adjustable parameter. This is the first component of the pseudo-biased variable. Under the batch update mechanism, the pre-set value is used first. The control quantity is calculated using other pseudo-partial derivative parameters, and then updated synchronously every N cycles using multi-time data. All parameters other than those mentioned above.

6. A distributed data-driven model-free adaptive output tracking control system for unmanned aerial vehicle swarms, used to implement the method as described in any one of claims 1-5, characterized in that, It includes a data acquisition module, a neighborhood error construction module, a batch parameter update module, a model-free adaptive control module, and a communication update module; The data acquisition module is used to acquire the location information of each follower drone and neighboring drones in real time using the sensors on the drone. The neighborhood error construction module is used to design a neighborhood output tracking error that integrates the local location tracking error and the location consistency error of neighboring UAVs. The batch parameter update module is used to perform batch estimation and update of pseudo-biased parameters every N sampling periods based on the assumption of slow time-varying UAV flight parameters. The model-free adaptive control module constructs a dynamic linearized data model based on position information and velocity command input data. It utilizes neighborhood output tracking error and pseudo-biased steering input to construct a distributed control law, calculates the velocity command input of each UAV, and achieves accurate position tracking of time-varying reference trajectory, thus realizing model-free adaptive control. The communication update module is used to send the position and speed information of the navigator drone to the follower drone.

Images