Resource-limited multi-unmanned aerial vehicle optimization backstepping formation fault-tolerant control method

Abstract

The invention relates to the technical field of unmanned aerial vehicle formation, in particular to a resource-limited multi-unmanned aerial vehicle optimization backstepping formation fault-tolerant control method, which comprises the following steps of: constructing a nonlinear dynamic model; determining a formation communication topological structure according to a formation task requirement; designing a neural network state observer; designing a trigger mechanism for state updating of the observer; calculating a formation tracking error and a tracking error of a formation global system; based on a backstepping control method, introducing first-order filtering at the same time, and calculating a speed tracking error; virtual control is introduced, and a filtering tracking error is calculated; designing an event triggering mechanism of an unmanned aerial vehicle controller based on the tracking error, the speed tracking error, the virtual control input and the first-order filtering of the global system; designing a neural network and optimizing a fault-tolerant controller; and the execution network and evaluation network architecture based on reinforcement learning optimizes control parameters. According to the invention, the response capability of the multi-unmanned aerial vehicle formation system after the unmanned aerial vehicles fail is significantly improved, and the operation efficiency of the formation system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) formations, and in particular to a fault-tolerant control method for an optimized backstepping formation of multiple UAVs with limited resources. Background Art

[0002] With the widespread application of multi-UAV systems in fields such as military reconnaissance, disaster relief, and logistics and transportation, their formation control technology faces increasingly severe challenges. In complex dynamic operating environments, traditional continuous-time control methods not only lead to a large number of redundant controller updates, resulting in a waste of computing resources, but also make it difficult to effectively respond to emergencies such as actuator failures and communication interruptions. Existing formation control strategies based on information exchange between adjacent UAVs are prone to chain propagation of fault information when a single UAV fails, seriously threatening the stability of the entire system. Especially when communication resources are limited, frequent information exchange not only increases network load, but also reduces the system's response speed to sudden failures, and in severe cases, it can even lead to the collapse of the entire formation system. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention provides a fault-tolerant control method for multi-UAV optimized backstepping formations with limited resources. Even if a UAV in the formation fails, the UAV formation can still operate according to the target trajectory. At the same time, the event trigger mechanism is integrated to reduce unnecessary updates of the controller, further improve the utilization efficiency of the formation resources, and provide a basis for the large-scale expansion of formation clusters.

[0004] The present invention is achieved through the following technical solutions: A fault-tolerant control method for optimized backstepping formation of multiple UAVs with limited resources is provided, which includes the following steps: S1. Construct a nonlinear dynamic model with actuator faults, disturbances and uncertainties; S2. Determine the UAV formation communication topology based on the Leader-Follower architecture and distributed control method according to the formation mission requirements; S3. Design a neural network state observer: in, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, represents the observer parameters that need to be designed, Indicates the system status output in the trigger state, represents the estimated value of the observer state output, represents the system state estimate of the observer, represents the estimated weights of the neural network, represents the basis function vector, represents the transpose of the matrix, represents the efficiency factor of the actuator, where , is the lower limit of the effectiveness factor; S4. Design a trigger mechanism for observer state update based on actual state information and observation information; S5. Calculate the UAV formation tracking error based on the consistent formation control protocol ; Based on the distributed communication protocol between adjacent UAVs, the tracking error of the formation global system is calculated ; S6, based on the backstepping control method and the introduction of first-order filtering , calculate the velocity tracking error ;Introduction of virtual control , calculate the filter tracking error ; S7, tracking error based on the global system , velocity tracking error , virtual control input and first-order filtering Design an event triggering mechanism for the drone controller; S8. Design a neural network and optimized fault-tolerant controllers ; S9. Execution Network Based on Reinforcement Learning and evaluation network Architecture optimization control parameters.

[0005] Furthermore, in step S1, a first nonlinear dynamic model is defined when a UAV failure occurs in the formation system: in, , , , represents the transpose of the matrix, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, represents the total uncertainty of the system, , Indicates a bias fault in the actuator, represents the external disturbance suffered, Indicates the speed of the drone; Based on the first nonlinear dynamic model, the system model is written as: in, Indicates location , Indicates speed , represents the output of the system, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, Represents the status output of the drone.

[0006] Furthermore, in step S2, the formation tracking error is first established ,in Indicates the status output of the drone, Indicates the mission information of the pilot drone. Indicates the formation information that needs to be maintained; according to the consistent formation control protocol, the tracking error of the global formation system is established ,in , Indicates the accumulation symbol, Represents the adjacency matrix The elements in Degree matrix Elements in; introduce first-order filtering And establish the velocity tracking error ; Establish a first-order filter tracking error ,in: represents the virtual control input in backstepping control, , represents the filtering parameters, and Represent the initial values ​​of virtual control and first-order filtering respectively.

[0007] Furthermore, in step S4, based on the observed drone state estimation value and the drone state trigger value obtained under the event trigger mechanism, a trigger mechanism is designed as follows: in, Indicates the system status output in the trigger state, Indicates the status output of the drone, Indicates the time when the last event was triggered. Indicates the time when the next event is triggered. represents the lower bound, that is, the maximum lower bound of the set of time instances, Indicates the trigger threshold that needs to be designed.

[0008] Furthermore, in step S7, the trigger mechanism is set as: in, Indicates the filter value in the trigger state, represents the filter value, Indicates the time when the last event was triggered. Indicates the time when the next event is triggered. represents the lower bound, that is, the maximum lower bound of the set of time instances, , and Indicates the trigger threshold that needs to be designed.

[0009] Furthermore, in step S8, considering the aggregate uncertainty in the formation system, a neural network is used for approximation to design a neural network weight adaptation law: in, represents the estimated value of the perturbed neural network, express The first time derivative of represents a positive constant gain, represents the basis function vector related to the formation error in the trigger state, Represents the transpose of a matrix; Designing optimized control inputs in triggering states in, represents a positive constant gain, represents the estimated value of the execution network weight, represents the basis function vector related to the formation error in the trigger state, Represents the estimated weights of the neural network Represents the transpose of a matrix.

[0010] Furthermore, in step S9, an adaptive law of the evaluation network is designed: Design the adaptive law of the execution network: in: represents the basis function vector related to the formation error in the trigger state, represents the neural network estimation value of the evaluation network, represents the neural network estimate of the execution network, express The first time derivative of express The first time derivative of 、 represents a positive constant gain, express dimensional identity matrix, Represents the transpose of a matrix.

[0011] Beneficial effects of the present invention: The fault-tolerant control method for optimized backstepping formation of multiple UAVs with limited resources provided by the present invention can ensure that, under the guidance of the lead UAV, even if one or more UAVs in the formation system fail, the UAV formation can still complete the communication of mission information based on the information transmission between adjacent UAVs and maintain a stable formation structure. At the same time, based on the designed trigger mechanism scheme, it is ensured that the designed observer and UAV controller can be updated on demand according to the system status.

[0012] The present invention can significantly improve the responsiveness of a multi-UAV formation system when a single UAV fails, effectively reducing formation system crashes or mission information loss due to UAV failures. It also ensures that controller information can be updated on demand, further improving the operational efficiency of the formation system. Experimental verification further demonstrates the reliability and robustness of this method in practical applications. This method not only has important theoretical significance but also demonstrates great potential in engineering applications. The designed controller enables on-demand updates, reduces resource consumption, and provides a foundation for the expansion of large-scale UAV formations, providing technical support and driving force for the rapid development of the UAV industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 Schematic diagram of the communication topology structure of a multi-UAV formation in the present invention.

[0014] Figure 2 This is the flight trajectory diagram of the multi-UAV formation in the present invention.

[0015] Figure 3 Schematic diagram of the quadrilateral formation layout of the multi-UAV formation in the present invention at 12s, 20s, and 30s.

[0016] Figure 4 This is a disturbance weight estimation curve diagram in the present invention.

[0017] Figure 5 This is the weight curve diagram of the evaluation network in the present invention.

[0018] Figure 6 This is a graph of the weight of the network executed in the present invention.

[0019] Figure 7 This is a tracking error curve diagram of the UAV formation in the present invention.

[0020] Figure 8 This is the X-axis control input curve diagram of the UAV formation in the present invention.

[0021] Figure 9 This is the Y-axis control input curve diagram of the UAV formation in the present invention.

[0022] Figure 10 This is the algorithm structure diagram of the present invention.

[0023] As shown in the figure: UAV0-leading drone, UAV1-following drone No. 1, UAV2-following drone No. 2, UAV3-following drone No. 3, UAV4-following drone No. 4. DETAILED DESCRIPTION

[0024] In order to clearly illustrate the technical features of this solution, this solution is described below through specific implementation methods.

[0025] like Figure 10 As shown in FIG, a fault-tolerant control method for optimizing backstepping formation of multiple UAVs with limited resources includes the following steps: S1. Construct a nonlinear dynamic model with actuator faults, disturbances and uncertainties; Define the first nonlinear dynamic model when a UAV failure occurs in the formation system: in, , , , represents the transpose of the matrix, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, represents the total uncertainty of the system, , Indicates a bias fault in the actuator, represents the external disturbance suffered, Indicates the speed of the drone; Based on the first nonlinear dynamic model, the system model is written as: in, Indicates location , Indicates speed , represents the output of the system, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, Represents the status output of the drone.

[0026] S2. According to the formation mission requirements, determine the UAV formation communication topology based on the Leader-Follower architecture and distributed control method.

[0027] First, establish the formation tracking error ,in Indicates the status output of the drone, Indicates the mission information of the pilot drone. Indicates the formation information that needs to be maintained; according to the consistent formation control protocol, the tracking error of the global formation system is established ,in , Indicates the accumulation symbol, Represents the adjacency matrix The elements in Degree matrix Elements in; introduce first-order filtering And establish the velocity tracking error ; Establish a first-order filter tracking error ,in: represents the virtual control input in backstepping control, , represents the filtering parameters, and Represent the initial values ​​of virtual control and first-order filtering respectively.

[0028] S3. Design a neural network state observer: in, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, represents the observer parameters that need to be designed, Indicates the system status output in the trigger state, represents the estimated value of the observer state output, represents the system state estimate of the observer, represents the estimated weights of the neural network, represents the basis function vector, represents the transpose of the matrix, represents the efficiency factor of the actuator, where , is the lower limit of the effectiveness factor; S4. Design a trigger mechanism for observer state update based on actual state information and observation information; Based on the observed drone state estimation value and the drone state trigger value obtained under the event trigger mechanism, the trigger mechanism is designed as follows: in, Indicates the system status output in the trigger state, Indicates the status output of the drone, Indicates the time when the last event was triggered. Indicates the time when the next event is triggered. represents the lower bound, that is, the maximum lower bound of the set of time instances, Indicates the trigger threshold that needs to be designed.

[0029] S5. Calculate the UAV formation tracking error based on the consistent formation control protocol ; Based on the distributed communication protocol between adjacent UAVs, the tracking error of the formation global system is calculated .

[0030] S6, based on the backstepping control method and the introduction of first-order filtering , calculate the velocity tracking error . Introducing virtual control , calculate the filter tracking error .

[0031] S7, tracking error based on the global system , velocity tracking error , virtual control input and first-order filtering Design an event triggering mechanism for the drone controller; The trigger mechanism is set as: in, Indicates the filter value in the trigger state, represents the filter value, Indicates the time when the last event was triggered. Indicates the time when the next event is triggered. represents the lower bound, that is, the maximum lower bound of the set of time instances, , and Indicates the trigger threshold that needs to be designed.

[0032] S8. Design a neural network and optimized fault-tolerant controllers ; Considering the aggregate uncertainty in the formation system, we use neural networks for approximation and design a neural network weight adaptation law: in, represents the estimated value of the perturbed neural network, express The first time derivative of represents a positive constant gain, represents the basis function vector related to the formation error in the trigger state, Represents the transpose of a matrix; Designing optimized control inputs in triggering states in, represents a positive constant gain, represents the estimated value of the execution network weight, represents the basis function vector related to the formation error in the trigger state, Represents the estimated weights of the neural network Represents the transpose of a matrix.

[0033] S9. Execution Network Based on Reinforcement Learning and evaluation network Architecture optimization control parameters.

[0034] Adaptive law of design evaluation network: Design the adaptive law of the execution network: in: represents the basis function vector related to the formation error in the trigger state, represents the neural network estimation value of the evaluation network, represents the neural network estimate of the execution network, express The first time derivative of express The first time derivative of 、 represents a positive constant gain, express dimensional identity matrix, Represents the transpose of a matrix.

[0035] The controller described above achieves semi-globally uniform bounded position and velocity tracking errors for the global formation system, while ensuring that the execution network, evaluation network, and perturbation weights remain bounded and ultimately converge to a constant value. Furthermore, the boundedness of the trigger error demonstrates that the designed trigger mechanism effectively avoids the Zeno phenomenon, preventing infinite triggering.

[0036] The specific verification steps are to use the analysis method based on Lyapunov stability to prove that when time tends to infinity, the distributed position tracking error and velocity tracking error converge asymptotically to 0 respectively.

[0037] The optimal fault-tolerant control method for multi-UAV formation failure proposed in this invention is used for simulation experiments, and the values ​​of the relevant parameters involved are as follows: Set the trajectory of the leader drone in the formation to , No. The formation vector of the UAVs is defined as , where the radius , angular velocity .

[0038] The actuator efficiency factor is defined as , the external disturbance is defined as: .

[0039] , , , , , The initial weight values ​​of the execution network and the evaluation network are set to , The simulation results are as follows: Figures 1 to 8 shown.

[0040] like Figure 1 As shown in the figure, a communication topology diagram for a multi-UAV formation is established using a pilot-follower and distributed control approach. The diagram includes four follower UAVs: UAV1 (follower UAV 1), UAV2 (follower UAV 2), UAV3 (follower UAV 3), and UAV4 (follower UAV 4), as well as a pilot UAV, UAV0. The pilot UAV can be configured as a virtual pilot. UAV1 and UAV4 receive mission information from the pilot UAV, and the four UAVs can exchange information with their neighbors.

[0041] like Figure 2-3 As shown, the formation drones are able to fly according to the predetermined trajectory and mission, and the position information at the initial stage, 12 seconds, 20 seconds and 30 seconds is intercepted. It can be clearly seen that the four drones are able to maintain a four-sided formation.

[0042] like Figure 4-6 As shown in the figure, the perturbation weight estimate, the weights of the execution network and the evaluation network can be guaranteed to be bounded. Since the state information of the execution network and the evaluation network is based on the trigger state, the convergence effect of the gradient descent can be clearly demonstrated as shown in the figure.

[0043] like Figure 7 As shown in the figure, the distributed tracking error in the formation system can achieve effective convergence, and after the fault is injected into the third UAV in the formation at the 15th second, the status of the four UAVs can be quickly stabilized.

[0044] like Figure 8-9 As shown, the actual control inputs of four UAVs and the control inputs under the trigger mechanism demonstrate that the proposed event triggering method effectively reduces many unnecessary controller signal updates.

[0045] Of course, the above description is not limited to the above examples. Technical features not described in the present invention can be achieved by or by adopting existing technologies, which will not be described here. The above embodiments and drawings are only used to illustrate the technical solutions of the present invention and are not limitations of the present invention. The present invention is described in detail with reference to the preferred implementation methods. Ordinary technicians in this field should understand that changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention do not depart from the purpose of the present invention and should also fall within the scope of protection of the claims of the present invention.

Claims

1. A fault-tolerant control method for multi-UAV optimized backstepping formation with limited resources, characterized by: The following steps are involved: S1. Construct a nonlinear dynamic model with actuator faults, disturbances and uncertainties; S2. Determine the UAV formation communication topology based on the Leader-Follower architecture and distributed control method according to the formation mission requirements; S3. Design a neural network state observer: in, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, represents the observer parameters that need to be designed, Indicates the system status output in the trigger state, represents the estimated value of the observer state output, represents the system state estimate of the observer, represents the estimated weights of the neural network, represents the basis function vector, represents the transpose of the matrix, represents the efficiency factor of the actuator, where , is the lower limit of the effectiveness factor; S4. Design a trigger mechanism for observer state update based on actual state information and observation information; S5. Calculate the UAV formation tracking error based on the consistent formation control protocol ; Based on the distributed communication protocol between adjacent UAVs, the tracking error of the formation global system is calculated ; S6, based on the backstepping control method and the introduction of first-order filtering , calculate the velocity tracking error ;Introduction of virtual control , calculate the filter tracking error ; S7, tracking error based on the global system , velocity tracking error , virtual control input and first-order filtering Design an event triggering mechanism for the drone controller; S8. Design a neural network and optimized fault-tolerant controllers ; S9. Execution Network Based on Reinforcement Learning and evaluation network Architecture optimization control parameters.

2. The method for fault-tolerant control of multi-UAV optimized backstepping formation with limited resources according to claim 1, characterized in that: In step S1, the first nonlinear dynamic model when a UAV failure occurs in the formation system is defined: in, , , , represents the transpose of the matrix, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, represents the total uncertainty of the system, , Indicates a bias fault in the actuator, represents the external disturbance suffered, Indicates the speed of the drone; Based on the first nonlinear dynamic model, the system model is written as: in, Indicates location , Indicates speed , represents the output of the system, and Respectively represent the first-order derivatives of position and velocity, represents the filtering control signal, Represents the status output of the drone.

3. The fault-tolerant control method for multi-UAV optimized backstepping formation with limited resources according to claim 1, characterized in that: In step S2, first establish the formation tracking error ,in Indicates the status output of the drone, Indicates the mission information of the pilot drone. Indicates the formation information that needs to be maintained; according to the consistent formation control protocol, the tracking error of the global formation system is established ,in , Indicates the accumulation symbol, Represents the adjacency matrix The elements in Degree matrix Elements in; introduce first-order filtering And establish the velocity tracking error ; Establish a first-order filter tracking error ,in: represents the virtual control input in backstepping control, , represents the filtering parameters, and Represent the initial values ​​of virtual control and first-order filtering respectively.

4. The fault-tolerant control method for multi-UAV optimized backstepping formation with limited resources according to claim 1, characterized in that: In step S4, based on the observed drone state estimation value and the drone state trigger value obtained under the event trigger mechanism, the observer trigger mechanism is designed as follows: in, Indicates the system status output in the trigger state, Indicates the status output of the drone, Indicates the time when the last event was triggered. Indicates the time when the next event is triggered. represents the lower bound, that is, the maximum lower bound of the set of time instances, Indicates the trigger threshold that needs to be designed.

5. The method for fault-tolerant control of multi-UAV optimized backstepping formation with limited resources according to claim 1, characterized in that: In step S7, the trigger mechanism is set as: in, Indicates the filter value in the trigger state, represents the filter value, Indicates the time when the last event was triggered. Indicates the time when the next event is triggered. represents the lower bound, that is, the maximum lower bound of the set of time instances, , and Indicates the trigger threshold that needs to be designed.

6. The method for fault-tolerant control of multi-UAV optimized backstepping formation with limited resources according to claim 1, characterized in that: In step S8, considering the aggregate uncertainty in the formation system, the neural network is used for approximation and the neural network weight adaptation law is designed: in, represents the estimated value of the perturbed neural network, express The first time derivative of represents a positive constant gain, represents the basis function vector related to the formation error in the trigger state, Represents the transpose of a matrix; Designing optimized control inputs in triggering states in, represents a positive constant gain, represents the estimated value of the execution network weight, represents the basis function vector related to the formation error in the trigger state, Represents the estimated weights of the neural network Represents the transpose of a matrix.

7. The method for fault-tolerant control of multi-UAV optimized backstepping formation with limited resources according to claim 1, characterized in that: In step S9, the adaptive law of the evaluation network is designed: Design the adaptive law of the execution network: in: represents the basis function vector related to the formation error in the trigger state, represents the neural network estimation value of the evaluation network, represents the neural network estimate of the execution network, express The first time derivative of express The first time derivative of 、 represents a positive constant gain, express -dimensional identity matrix, Represents the transpose of a matrix.

Images