A data-driven air-ground cross-domain unmanned aerial vehicle switching robust model predictive control method and system

Abstract

The application provides a data-driven air-ground cross-domain unmanned aerial vehicle switching robust model predictive control method and system, and belongs to the air-ground cross-domain unmanned aerial vehicle attitude control field. In order to solve the problem that the existing air-ground cross-domain unmanned aerial vehicle control method is highly dependent on the precise mathematical model of the system established through mechanism analysis or experimental identification, the process is complicated and sensitive to system changes; the existing related data-driven model predictive control method is difficult to process data containing noise and continuous interference in online operation, resulting in the problem that the system lacks robustness. The application provides a switching model predictive control method based on offline input-state data containing noise, solves the minimum-maximum model predictive control optimization problem online to obtain the control instruction to ensure the safety and robustness of the unmanned aerial vehicle movement, and directly designs the control law through the offline collected input-state data, bypasses the explicit system identification step, and greatly reduces the air-ground cross-domain unmanned aerial vehicle controller design cost.

Description

Technical Field

[0001] This invention relates to the field of attitude control technology for air-to-ground cross-domain unmanned aerial vehicles (UAVs), and more specifically, to a data-driven robust model predictive control method and system for switching air-to-ground cross-domain UAVs. Background Technology

[0002] Air-to-ground cross-domain unmanned aerial vehicles (UAVs) demonstrate significant advantages in complex mission scenarios such as disaster relief, environmental monitoring, and underground inspection due to their combination of rapid aerial maneuverability and precise ground operations. These UAVs can flexibly switch between flight and ground movement modes to adapt to mission requirements in different environments. During the transition between air and ground modes, the number of state variables in the UAV system changes: the aerial phase is typically described using six degrees of freedom, while the ground phase degenerates into planar motion. This variable-dimensional characteristic makes it impossible to uniformly describe the dynamic behavior of air-to-ground cross-domain UAVs using traditional fixed-dimensional differential or difference equations, thus making it difficult to directly apply traditional UAV control system design methods.

[0003] Currently, control methods for cross-domain air-to-ground UAVs largely rely on model-based frameworks. This involves first establishing a precise mathematical model of the system through mechanistic analysis or experimental identification, and then designing a controller based on this model. Such methods require designers to possess sufficient prior knowledge and parameter identification capabilities, making the process cumbersome and sensitive to system variations. To reduce reliance on precise models, data-driven control methods, which have gained significant attention in recent years, have emerged as a promising alternative. These methods can directly design controllers using offline-collected data without explicitly establishing a system model. However, most existing data-driven methods assume that the system has a non-singular, fixed-dimensional state-space representation, failing to address the state dimension jumps that occur during mode switching in cross-domain air-to-ground UAVs. In fact, this variable-dimensionality characteristic can be described by switching singular systems. In such systems, each subsystem is allowed to have algebraic constraints, and different subsystems can have dynamic variables of different dimensions, thus effectively characterizing structural abrupt changes during mode switching. Unfortunately, however, data-driven control methods for switching singular systems remain an open problem.

[0004] More importantly, air-to-ground cross-domain UAVs are inevitably affected by uncertainties such as external wind disturbances, uneven ground, and sensor noise in actual operating environments. Simultaneously, ensuring UAV safety and optimizing its control performance are crucial during the control process. Robust model predictive control (MMDC), with its ability to explicitly handle physical constraints and uncertain disturbances and optimize performance costs online, has become an ideal choice for controlling such systems. However, current robust MDC methods for switching singular systems still primarily employ exact model-driven approaches. Furthermore, existing limited research on data-driven MDC for singular systems assumes that the acquired data is free of noise and the system is unaffected by any disturbances; these idealized assumptions severely limit the applicability of the methods in real-world scenarios. Therefore, there is an urgent need to develop robust MDC methods for switching singular systems driven by noisy data. Summary of the Invention

[0005] The technical problem to be solved by this invention is:

[0006] To address the problem that existing control methods for cross-domain UAVs rely heavily on establishing accurate mathematical models of the system through mechanistic analysis or experimental identification, which is a cumbersome process and sensitive to system changes; and that existing data-driven model predictive control methods struggle to handle noisy data and continuous disturbances during online operation, resulting in a lack of system robustness.

[0007] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0008] This invention provides a data-driven robust model predictive control method for air-to-ground cross-domain UAV handover, comprising the following steps:

[0009] S100. Based on the switched singular system model, define the system input and state variables, determine the singular matrix and disturbance matrix of each switched mode, determine the upper bound of the disturbance, and establish system constraints.

[0010] S200: Offline acquisition of system input-state data under various modes; based on the upper bound of the disturbance, the unknown system matrix of each mode is represented as a set form compatible with the offline data.

[0011] S300, the design considers the robust performance optimization objective of the air-to-ground cross-domain unmanned aerial vehicle system under continuous external interference, and establishes a mode-dependent minimum-maximum model predictive control optimization problem;

[0012] S400: The optimization problem constructed in step S300 is transformed into a numerically solvable convex optimization problem. During online operation, the corresponding convex optimization problem is solved according to the current mode and state, and the obtained control commands are applied to the air-to-ground cross-domain UAV.

[0013] Further, in step S100, the following are included:

[0014] Consider the following model of a switching singular system:

[0015] (1)

[0016] in, Represents system state variables; Indicates the control input quantity; Indicates external interference; Indicates the dimension of the system state variables; Indicates the dimension of the control input; Indicates the dimension of external interference; Indicates the sampling time; Indicates a switching signal, where Indicates aerial mode, Represents ground modes; , , and They represent the first Singular matrices, system matrix, input matrix, and disturbance matrix under different modes; conditions Established, among which For the first The dynamic order of the system under modal conditions;

[0017] Will Defined as a state variable, Defined as a control input, where , and These represent the roll angle, pitch angle, and yaw angle, respectively. , and These represent the roll rate, pitch rate, and yaw rate in the body coordinate system, respectively. , and These represent the input torques along the x, y, and z axes in the body coordinate system, respectively.

[0018] The continuous external disturbances experienced by the system are bounded, and their range is as follows:

[0019] (2)

[0020] in, Indicates the range of interference constraints; Denotes the known upper bound of the interference norm 2;

[0021] Introduce the following modally dependent system state and control input constraints:

[0022] (3)

[0023] (4)

[0024] in, and They represent the first The system state variables and control input variables of the modal system are constrained by a polyhedral matrix. and They represent the first The modal system state variables and control input variables have polyhedral constraints, among which and The dimensions are respectively and .

[0025] Further, in step S200, the following are included:

[0026] Offline acquisition for each modality Data ,in and They represent the first Modal 1 System state variables and control input variables This represents the corresponding state response; the corresponding data interference is represented as... The above data is organized into the following matrix form:

[0027] (5)

[0028] in, A data matrix representing the system state; A data matrix representing the control input; The data matrix representing external interference; A data matrix representing the state response;

[0029] According to the singular value decomposition method, the left and right transformation matrices are obtained for each switching mode. and :

[0030] (6)

[0031] in, Represents the transformed th The system matrix group under modal conditions, with subscripts 11, 12, 21, and 22, respectively, representing the first row and first column, the first row and second column, the second row and first column, and the second row and second column; Indicates the first Input matrix of the difference numerator system for singular systems under modal conditions; Indicates the first Disturbance matrix of the difference molecule system in a singular system under modal conditions; express 1-dimensional identity matrix

[0032] Introducing the transformed state The original switching singular system is transformed into a difference numerator system and an algebraic subsystem:

[0033] (7)

[0034] in, and Let the system states of the difference numerator system and the algebraic subsystem be represented respectively. and These represent the system matrices of the difference numerator system and the algebraic subsystem, respectively;

[0035] Based on the upper bound of the interference, the first... The actual system matrix and input matrix of the differential molecular system under modal conditions The set of data that belongs to the following :

[0036] (8)

[0037] in, Let be the kernel matrix of this set. , and They represent the first The state data matrix, state response data matrix of the modal differential molecular system, and state response matrix of the algebraic subsystem;

[0038] No. The system matrix of the real algebraic subsystem under modal conditions is obtained directly from the data:

[0039] (9)

[0040] Among them, symbols Represents the pseudo-inverse of a matrix;

[0041] The system matrix of the real switching exotic system The set of data that belongs to the following :

[0042] (10).

[0043] Further, in step S300, the following are included:

[0044] Optimize the following objectives online while ensuring system constraints are met. :

[0045] (11)

[0046] in, , and They represent Predictions at any time The system state, control inputs, and external disturbances at any given time. express Predictions at any time The system state at a given time is numerically consistent with the current state; for modal... , and These represent the weight matrices for the state variables and the input variables, respectively. A weighted scalar representing the ability to resist interference;

[0047] Construct a modally dependent min-max model predictive control optimization problem:

[0048] (12).

[0049] Further, in step S400, the following is included:

[0050] To ensure the numerical solvability of the optimization problem, the control sequence is parameterized into the following state feedback form:

[0051] (13)

[0052] in, State feedback gain;

[0053] The first The model predictive control optimization problem under modal conditions is transformed into the following numerically solvable convex optimization problem:

[0054] (14)

[0055] in, These are decision variables used to ensure system stability; These are decision variables used to synthesize control gains; These are decision variables used to process data; It is the upper limit of the cost; These are the state weights of the differential molecular system after transformation; symbol This indicates that the vector is in the first position. The nth element or the nth element of the matrix OK;

[0056] Based on the current system mode and current state Solving the convex optimization problem to synthesize state feedback control gain in real time. The following control variables will be applied to the controlled object:

[0057] (15).

[0058] A data-driven robust model predictive control system for air-to-ground cross-domain UAV handover, the system having program modules corresponding to the above steps, and executing the steps in the above-described data-driven robust model predictive control method for air-to-ground cross-domain UAV handover during runtime.

[0059] A computer-readable storage medium storing a computer program configured to, when invoked by a processor, implement the steps of a data-driven, air-to-ground cross-domain unmanned aerial vehicle (UAV) handover robust model predictive control method.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] This invention proposes a data-driven attitude control method for air-to-ground cross-domain UAVs. It directly designs control laws using offline-collected input-state data, bypassing explicit system identification steps and greatly reducing the design cost of air-to-ground cross-domain UAV controllers.

[0062] This invention takes into account the noise impact on offline data acquisition and the continuous interference during online operation, and integrates a robust control framework to improve the robustness of air-to-ground cross-domain UAVs under complex working conditions.

[0063] This invention presents a switching model predictive control method for air-to-ground cross-domain unmanned aerial vehicles (UAVs) based on a switching singular system design. It optimizes the system's stability and anti-interference capabilities online through a rolling time domain approach, while ensuring the safety of the UAV during execution. Attached Figure Description

[0064] Figure 1 This is a flowchart of a data-driven robust model predictive control method for air-to-ground cross-domain UAV handover in an embodiment of the present invention;

[0065] Figure 2 This is a structural diagram of the air-to-ground cross-domain unmanned aerial vehicle system in an embodiment of the present invention;

[0066] Figure 3 This is a signal switching curve diagram for air-to-ground cross-domain UAVs in an embodiment of the present invention;

[0067] Figure 4 This is a diagram showing the attitude angle response curve of an air-to-ground cross-domain UAV in an embodiment of the present invention.

[0068] Figure 5 This is a graph showing the attitude angular velocity response of an air-to-ground cross-domain UAV in an embodiment of the present invention.

[0069] Figure 6 This is a control input curve diagram for a cross-domain UAV in an embodiment of the present invention. Detailed Implementation

[0070] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0071] Specific Implementation Plan 1: Combining Figure 1 As shown, this invention provides a data-driven robust model predictive control method for air-to-ground cross-domain UAV handover, comprising the following steps:

[0072] S100. Based on the switched singular system model, define the system input and state variables, determine the singular matrix and disturbance matrix of each switched mode, determine the upper bound of the disturbance, and establish system constraints.

[0073] Consider the following model of a switching singular system:

[0074] (1)

[0075] in, Represents system state variables; Indicates the control input quantity; Indicates external interference; Indicates the dimension of the system state variables; Indicates the dimension of the control input; Indicates the dimension of external interference; Indicates the sampling time; Indicates a switching signal, where Indicates aerial mode, Represents ground modes; , , and They represent the first Singular matrices, system matrix, input matrix, and disturbance matrix under different modes; conditions Established, among which For the first The dynamic order of the system under modal conditions;

[0076] Will Defined as a state variable, Defined as a control input, where , and These represent the roll angle, pitch angle, and yaw angle, respectively. , and These represent the roll rate, pitch rate, and yaw rate in the body coordinate system, respectively. , and These represent the input torques along the x, y, and z axes in the body coordinate system; the singular matrix in the constructed air-to-ground cross-domain UAV switching singular system. It does not contain system parameter information, which can be known a priori; the interference matrix. The channel representing the interference effect can therefore be known a priori; and The matrix contains a large amount of model information that is unknown.

[0077] The continuous external disturbances experienced by the system are bounded, and their range is as follows:

[0078] (2)

[0079] in, Indicates the range of interference constraints; Denotes the known upper bound of the interference norm 2;

[0080] Introduce the following modally dependent system state and control input constraints:

[0081] (3)

[0082] (4)

[0083] in, and They represent the first The system state variables and control input variables of the modal system are constrained by a polyhedral matrix. and They represent the first The modal system state variables and control input variables have polyhedral constraints, among which and The dimensions are respectively and ;

[0084] S200: Offline acquisition of system input-state data under various modes; based on the upper bound of the disturbance, the unknown system matrix of each mode is represented as a set form compatible with the offline data.

[0085] For each modality offline acquisition A data point, represented as ,in and They represent the first Modal 1 System state variables and control input variables This represents the corresponding state response; the corresponding data interference is represented as... It strictly satisfies constraint (2); the above data is organized into the following matrix form:

[0086] (5)

[0087] in, A data matrix representing the system state; A data matrix representing the control input; The data matrix representing external interference; A data matrix representing the state response;

[0088] According to the singular value decomposition method, the left and right transformation matrices can be obtained for each switching mode. and Make:

[0089] (6)

[0090] in, Represents the transformed th The system matrix group under modal conditions, with subscripts 11, 12, 21, and 22, respectively, representing the first row and first column, the first row and second column, the second row and first column, and the second row and second column; Indicates the first Input matrix of the difference numerator system for singular systems under modal conditions; Indicates the first Disturbance matrix of the difference molecule system in a singular system under modal conditions; express An identity matrix of 3D;

[0091] Therefore, by introducing the transformed state The original switching singular system (1) can be transformed into a difference numerator system and an algebraic subsystem:

[0092] (7)

[0093] in, and Let the system states of the difference numerator system and the algebraic subsystem be represented respectively. and These represent the system matrices of the difference numerator system and the algebraic subsystem, respectively;

[0094] Based on the upper bound of the interference, the first... The actual system matrix and input matrix of the differential molecular system under modal conditions The set of data that belongs to the following :

[0095] (8)

[0096] in, Let be the kernel matrix of this set. , and They represent the first The state data matrix, state response data matrix of the modal differential molecular system, and state response matrix of the algebraic subsystem;

[0097] No. The actual system matrix of the algebraic subsystem under modal conditions can be obtained directly from the data:

[0098] (9)

[0099] Among them, symbols Represents the pseudo-inverse of a matrix;

[0100] Therefore, the system matrix of a real switching singular system The set of data that belongs to the following :

[0101] (10)

[0102] S300, design consideration of robust performance optimization objectives for air-to-ground cross-domain unmanned aerial vehicle system with continuous interference, and establish mode-dependent minimum-maximum model predictive control optimization problem;

[0103] To ensure the robustness of the system, the following objectives will be optimized online while maintaining system constraints. :

[0104] (11)

[0105] in, , and They represent Predictions at any time The system state, control inputs, and external disturbances at any given time. express Predictions at any time The system state at a given time is numerically consistent with the current state; for modal... , and These represent the weight matrices for the state variables and the input variables, respectively. A weighted scalar representing the ability to resist interference;

[0106] Therefore, the following modally dependent min-max model predictive control optimization problem can be constructed:

[0107] (12)

[0108] S400: Transform the optimization problem constructed in step S300 into a numerically solvable convex optimization problem. During online operation, solve the corresponding optimization problem based on the current mode and state, and apply the obtained control commands to the controlled object.

[0109] To ensure that the optimization problem is numerically solvable, the control sequence is parameterized into the following state feedback form:

[0110] (13)

[0111] in, State feedback gain;

[0112] Then, the first The model predictive control optimization problem (12) under modal conditions is transformed into the following numerically solvable convex optimization problem:

[0113] (14)

[0114] in, These are decision variables used to ensure system stability; These are decision variables used to synthesize control gains; These are decision variables used to process data; It is the upper limit of the cost; These are the state weights of the differential molecular system after transformation; symbol This indicates that the vector is in the first position. The nth element or the nth element of the matrix OK;

[0115] Finally, based on the current system modes and current state Solve the convex optimization problem (14) to control the gain using real-time synthesized state feedback. The following control variables will be applied to the controlled object:

[0116] (15)

[0117] Specific Implementation Scheme 2: The present invention provides a data-driven robust model predictive control system for air-to-ground cross-domain UAV handover. This system has program modules corresponding to the above steps, and executes the steps in the above-described data-driven robust model predictive control method for air-to-ground cross-domain UAV handover during runtime.

[0118] The other combinations and connections in this implementation scheme are the same as in Specific Implementation Scheme 1.

[0119] Specific Implementation Scheme 3: The present invention provides a computer-readable storage medium storing a computer program configured to implement, when called by a processor, the steps of a data-driven air-to-ground cross-domain UAV switching robust model predictive control method.

[0120] The other combinations and connections in this implementation scheme are the same as in Specific Implementation Scheme 1.

[0121] Example

[0122] In this embodiment, consider as follows Figure 2 The attitude dynamics of the air-to-ground cross-domain UAV shown is expressed in the air mode as shown in Equation (16):

[0123] (16)

[0124] Its attitude dynamics in the ground mode are expressed as shown in equation (17):

[0125] (17)

[0126] in, = 0.319 kg·m 2 , = 0.256 kg·m 2 and = 0.352 kg·m 2 These are the moments of inertia of the x, y, and z axes in the body coordinate system, respectively. = 0.06、 = 0.06 and = 0.02 represent the air resistance coefficients of the UAV's x, y, and z axes in the body coordinate system, respectively; , and These represent the disturbance torques acting on the x, y, and z axes in the body coordinate system, respectively. Indicates the first Equilibrium vector of UAV in different modes;

[0127] In this embodiment, the attitude system of the discretized air-to-ground cross-domain UAV is controlled. All system parameters are assumed to be unknown, the equilibrium point vector is set as the origin, and the sampling period is 0.02 s.

[0128] Based on step S100, a switching singular system model is introduced:

[0129] (18)

[0130] Define the state variable as The control input is defined as follows: Interference quantity is defined as The singular matrix and interference matrix in the air mode are respectively and The singular matrix and disturbance matrix in the ground mode are respectively and The upper limit of the disturbance is set at 0.01, and the attitude angle range is determined to be... rad, attitude angular velocity range is rad / s, triaxial input torque range is N·m.

[0131] According to step S200, an initial state within the state constraint range is randomly given, and a random uniformly distributed input sequence of length 200 is given to obtain offline data of length 200. The unknown system matrix of each mode is represented as a set form compatible with the offline data according to the method in Specific Implementation Method 1.

[0132] According to step S300, a robust performance optimization objective for an air-to-ground cross-domain unmanned aerial vehicle system considering continuous interference is designed, with its weight parameters set as follows: , , And based on the method in Specific Implementation Method 1, a modality-dependent min-max model predictive control optimization problem is established.

[0133] According to step S400, the constructed optimization problem is transformed into a numerically solvable convex optimization problem. During online execution, the corresponding optimization problem is solved based on the current mode and state, and the obtained control commands are applied to the controlled object. The air-to-ground mode switching signal for the air-to-ground cross-domain UAV is as follows: Figure 3 As shown in the figure. This embodiment compares the method proposed in this invention with a data-driven state feedback control method and a data-driven model predictive control method based on Willems' lemma. The UAV attitude angle response curve, UAV attitude angular velocity response curve, and UAV input torque curve are shown in the figure. Figure 4 , Figure 5 and Figure 6 As shown, the performance cost of the UAV's actual execution in the first 200 steps was calculated, and the performance costs of the three methods were 11.48, 13.35, and 15.10, respectively. The results show that the method proposed in this invention can guarantee the robust stability and operational safety of the air-to-ground cross-domain UAV under air-to-ground mode switching, and its control performance is superior to previous methods.

[0134] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A data-driven robust model predictive control method for air-to-ground cross-domain UAV handover, characterized in that, Includes the following steps: S100. Based on the switched singular system model, define the system input and state variables, determine the singular matrix and disturbance matrix of each switched mode, determine the upper bound of the disturbance, and establish system constraints. S200: Offline acquisition of system input-state data under various modes; based on the upper bound of the disturbance, the unknown system matrix of each mode is represented as a set form compatible with the offline data. S300, the design considers the robust performance optimization objective of the air-to-ground cross-domain unmanned aerial vehicle system under continuous external interference, and establishes a mode-dependent minimum-maximum model predictive control optimization problem; S400: The optimization problem constructed in step S300 is transformed into a numerically solvable convex optimization problem. During online operation, the corresponding convex optimization problem is solved according to the current mode and state, and the obtained control commands are applied to the air-to-ground cross-domain UAV.

2. The data-driven robust model predictive control method for air-to-ground cross-domain UAV handover according to claim 1, characterized in that: In step S100, the following are included: Consider the following model of a switching singular system: (1) in, Represents system state variables; Indicates the control input quantity; Indicates external interference; Indicates the dimension of the system state variables; Indicates the dimension of the control input; Indicates the dimension of external interference; Indicates the sampling time; Indicates a switching signal, where Indicates aerial mode, Represents ground modes; , , and They represent the first Singular matrices, system matrix, input matrix, and disturbance matrix under different modes; conditions Established, among which For the first The dynamic order of the system under modal conditions; Will Defined as a state variable, Defined as a control input, where , and These represent the roll angle, pitch angle, and yaw angle, respectively. , and These represent the roll rate, pitch rate, and yaw rate in the body coordinate system, respectively. , and These represent the input torques along the x, y, and z axes in the body coordinate system, respectively. The continuous external disturbances experienced by the system are bounded, and their range is as follows: (2) in, Indicates the range of interference constraints; Denotes the known upper bound of the interference norm 2; Introduce the following modally dependent system state and control input constraints: (3) (4) in, and They represent the first The system state variables and control input variables of the modal system are constrained by a polyhedral matrix. and They represent the first The modal system state variables and control input variables have polyhedral constraints, among which and The dimensions are respectively and .

3. The data-driven robust model predictive control method for air-to-ground cross-domain UAV handover according to claim 2, characterized in that: Step S200 includes, Offline acquisition for each modality Data ,in and They represent the first Modal 1 Each system state variable and control input variable This represents the corresponding state response; the corresponding data interference is represented as... Organize the above data into the following matrix form: (5) in, A data matrix representing the system state; A data matrix representing the control input; The data matrix representing external interference; A data matrix representing the state response; According to the singular value decomposition method, the left and right transformation matrices are obtained for each switching mode. and : (6) in, Represents the transformed th The system matrix group under modal conditions, with subscripts 11, 12, 21, and 22, respectively, representing the first row and first column, the first row and second column, the second row and first column, and the second row and second column; Indicates the first Input matrix of the difference numerator system for singular systems under modal conditions; Indicates the first Disturbance matrix of the difference molecule system in a singular system under modal conditions; express An identity matrix of dimensionality; Introducing the transformed state The original switching singular system is transformed into a difference numerator system and an algebraic subsystem: (7) in, and Let the system states of the difference numerator system and the algebraic subsystem be represented respectively. and These represent the system matrices of the difference numerator system and the algebraic subsystem, respectively; Based on the upper bound of the interference, the first... The actual system matrix and input matrix of the differential molecular system under modal conditions The set of data that belongs to the following : (8) in, Let be the kernel matrix of this set. , and They represent the first The state data matrix, state response data matrix of the modal differential molecular system, and state response matrix of the algebraic subsystem; No. The system matrix of the real algebraic subsystem under modal conditions is obtained directly from the data: (9) Among them, symbols Represents the pseudo-inverse of a matrix; The system matrix of the real switching exotic system The set of data that belongs to the following : (10)。 4. The data-driven robust model predictive control method for air-to-ground cross-domain UAV handover according to claim 3, characterized in that: Step S300 includes, Optimize the following objectives online while ensuring system constraints are met. : (11) in, , and They represent Predictions at any time The system state, control inputs, and external disturbances at any given time. express Predictions at any time The system state at a given time is numerically consistent with the current state; for modal... , and These represent the weight matrices for the state variables and the input variables, respectively. A weighted scalar representing the ability to resist interference; Construct a modally dependent min-max model predictive control optimization problem: (12)。 5. The data-driven robust model predictive control method for air-to-ground cross-domain UAV handover according to claim 4, characterized in that: In step S400, the following are included: To ensure the numerical solvability of the optimization problem, the control sequence is parameterized into the following state feedback form: (13) in, For state feedback gain; The first The model predictive control optimization problem under modal conditions is transformed into the following numerically solvable convex optimization problem: (14) in, These are decision variables used to ensure system stability; These are decision variables used to synthesize control gains; These are decision variables used to process data; It is the upper bound of the cost; These are the state weights of the differential molecular system after transformation; symbol This indicates that the vector is in the first position. The nth element or the nth element of the matrix OK; Based on the current system mode and current state Solving the convex optimization problem to synthesize state feedback control gain in real time. The following control variables will be applied to the controlled object: (15)。 6. A data-driven robust model predictive control system for air-to-ground cross-domain UAV handover, characterized in that: The system has a program module corresponding to the steps of any one of the claims 1-5 above, and executes the steps in the above-described data-driven robust model predictive control method for air-to-ground cross-domain UAV handover when running.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program configured to, when invoked by a processor, implement the steps of the data-driven air-to-ground cross-domain unmanned aerial vehicle (UAV) handover robust model predictive control method according to any one of claims 1-5.

Images