An event-triggered air-ground dual-purpose unmanned aerial vehicle semi-Markov smooth model prediction control method and system

Abstract

The application discloses an event-triggered air-ground dual-purpose unmanned aerial vehicle semi-Markov smooth model prediction control method and system, belongs to the technical field of control theory and control engineering, and solves the problems of input jitter and high calculation amount of the air-ground dual-purpose unmanned aerial vehicle control.The method comprises the following steps: a dynamic event-triggered model prediction smooth control framework is proposed; a transition-dependent smooth controller and a mode-dependent stabilizing controller are designed; and the optimal control of the system in a disturbed environment is realized.Through state observation and a dynamic triggering mechanism, a model prediction control law is directly designed, complex online optimization calculation is not needed at each time, and efficient control of the random switching system can be realized.The application is suitable for the control of complex engineering systems such as unmanned aerial vehicles and power systems.

Description

Technical Field

[0001] This application relates to the fields of control theory and control engineering, and in particular to smooth model predictive control technology for computationally limited semi-Markov jump systems. Background Technology

[0002] In the design of air-to-ground unmanned aerial vehicles (UAVs), to balance ground mobility and aerial maneuverability, support structures such as wheel hubs are typically integrated on both sides of the fuselage. This structural distribution results in a system moment of inertia much larger than that of conventional multi-rotor UAVs. When the UAV encounters time-varying airflow disturbances or is under different load conditions, its dynamic characteristics will randomly switch between multiple specific linear sub-models. This process can be modeled as a semi-Markov jump system. Furthermore, since the system maintains a specific motion mode for a maximum physical time limit in reality, it is usually assumed that the dwell time has an upper bound. Therefore, such air-to-ground UAVs can be modeled as semi-Markov jump systems with an upper bounded dwell time, posing a significant challenge to the design of control methods.

[0003] On the other hand, dual-use air-to-ground UAV platforms face two challenges in control: First, the airframe's limited computing power by the embedded processing unit often makes it difficult for the controller to support high-frequency, long-term optimization calculations, directly resulting in significant computational delays during algorithm execution. Second, due to limitations in airframe structure and motor lifespan, control input jitter must be minimized during controller design. These characteristics make the control of such systems far more difficult than in conventional scenarios.

[0004] To address the above issues, Event-Triggered Model Predictive Control (ETMPC) has garnered widespread attention due to its superior capabilities in handling multivariable constraints, online optimization, and conserving computational resources. Some studies have extended ETMPC to Markov systems, reducing controller computational costs while maintaining system control performance through the design of triggering conditions and strategies. It must be noted that these methods essentially largely follow traditional event-triggered strategies, focusing on reducing communication transmission load while neglecting a key characteristic of MPC systems: the computational time required for online iterative optimization is often far greater than that required to solve for the control input with a known gain. This may lead to methods based on existing triggering mechanisms requiring frequent triggering to meet system performance requirements, increasing computational complexity. Furthermore, existing research primarily focuses on the steady-state performance of the system, neglecting its transient performance. Especially in switching systems incorporating event-triggered mechanisms, control input jitter occurs not only during mode switching but also during event triggering. This dual jitter phenomenon can lead to actuator damage or performance degradation; therefore, event-triggered switching systems have a more urgent need for smooth control mechanisms than conventional switching systems. Summary of the Invention

[0005] The purpose of this invention is to solve the technical problems of jitter in the control input and high computational load of air-to-ground dual-purpose UAVs, and to provide a semi-Markov smooth model predictive control method and system for air-to-ground dual-purpose UAVs based on event triggering.

[0006] The technical solution adopted by this invention to solve the above problems is: a semi-Markov smooth model predictive control method for an event-triggered air-to-ground dual-purpose UAV, the method comprising:

[0007] Step 1: Establish a discrete-time UAV system model based on a semi-Markov random process, incorporating external disturbances and multi-mode switching characteristics; determine the initial state and initial modes of the system; set parameters that satisfy the dynamic event triggering mechanism, and select the weight matrix and performance index constants for model predictive control;

[0008] Step 2: At each sampling time, obtain the current system state; calculate the difference error between the current state and the state at the previous trigger time, and make a judgment based on the preset dynamic event trigger conditions in combination with auxiliary dynamic variables; if the trigger conditions are met or the current time is the initial time, it is determined to be triggered, and proceed to Step 3 to update the control law; if the trigger conditions are not met, it is determined to be not triggered, skip the optimization solution process, and directly proceed to Step 4.

[0009] Step 3: When an event is triggered, based on the current system modes and states, construct an infinite time-domain model predictive control optimization problem that includes smoothness constraints and stability constraints; solve the optimization problem online using linear matrix inequality techniques, calculate the positive definite matrix and decision variables that minimize the cost function, and thus obtain the updated smooth feedback control gain and trigger matrix to ensure that the system satisfies recursive feasibility and mean square input state stability, while limiting the jitter amplitude of the control input;

[0010] Step 4: If the system is in the transition phase after mode switching or event triggering, select a transition-dependent smooth controller to suppress control input jitter; if the system is in the stable phase of modal operation, select a mode-dependent stabilizing controller to maintain system performance; use the selected controller and the current system state to calculate the current control input.

[0011] Step 5: Send the calculated control input to the actuator to apply to the UAV system, driving the system state to update to the next moment; then the system runs in closed loop, jumps back to step 2, and performs the event trigger judgment and control loop for the next moment until the controller design is completed or the task is completed.

[0012] Furthermore, in step one, the unmanned aerial vehicle (UAV) system model is as follows:

[0013]

[0014] in, Sampling time, For system status, These are platform speed, pitch angle, acceleration, and pitch angular velocity, respectively. To control the input, External disturbances For system modes, For a set of subsystems, It is a system mode The system matrix, It is a system mode The input matrix, It is a system mode The perturbation matrix;

[0015] System Matrix

[0016]

[0017] Input matrix

[0018]

[0019] in, , Indicates the equilibrium point parameters. The sampling period is For the body mass, It is the acceleration due to gravity. For rotational inertia, For the body size, The friction coefficients under different modes.

[0020] Furthermore, step one includes: setting an upper bound on the residence time, a weight matrix, and input constraints. upper boundary of disturbance Triggering algorithm related parameters And the system's semi-Markov kernel.

[0021] Furthermore, the triggering condition in step two is:

[0022]

[0023] in, The set of natural numbers, This refers to the nth switch after the tth trigger time. Represents the time of the t-th trigger. ,make , This represents the difference between the current system state and the system state at the time of the most recent trigger. , This serves as the trigger matrix for subsequent solutions. As an auxiliary variable,

[0024] .

[0025] Furthermore, the positive definite matrix in step three is: ,right The Lyapunov function is defined as follows:

[0026]

[0027] in, .

[0028] Furthermore, the optimization problem in step three specifically includes: At time t, for a given weight matrix Q, R, There exists a positive definite matrix. sum matrix This makes the cost function:

[0029]

[0030]

[0031] in , , , for Natural numbers within, The upper limit of the dwell time, It is a half-Markov nucleus. For the Lyapunov matrix related to the length of stay, These represent the system's stability performance, smoothing performance, and disturbance suppression performance, respectively. The system matrix is ​​the system mode r. Let r be the input matrix of the system mode r; Let be the perturbation matrix of system mode r;

[0032]

[0033] in , , , and These are the invariant set matrices within the subsystem and at the mode switching time, respectively, and both are positive definite. It is a positive definite matrix. Representation matrix The Line number Elements at column, Indicate input constraints The One element;

[0034]

[0035] in , for Time controller Control gain , .

[0036] Furthermore, in step four, the calculation method for the control input is as follows:

[0037] Based on the current system state, according to the formula Select the control gain to use and calculate the control input.

[0038] .

[0039] Secondly, the present invention provides a method system for event-triggered air-to-ground dual-purpose UAV semi-Markov smooth model predictive control. The system has a program module corresponding to the steps of the event-triggered air-to-ground dual-purpose UAV semi-Markov smooth model predictive control method described above, and executes the steps in the event-triggered air-to-ground dual-purpose UAV semi-Markov smooth model predictive control method during runtime.

[0040] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, it executes the steps of the event-triggered semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV as described above.

[0041] Fourthly, the present invention provides a computer-readable storage medium for storing a computer program that executes a semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV based on event triggering, as described above.

[0042] The beneficial effects of this invention are:

[0043] Specifically, this manifests in the following aspects:

[0044] (1) A smooth control strategy for the "double jitter" problem in randomly switched systems is proposed, which significantly improves the transient performance and actuator lifespan of the system. Addressing the shortcomings of existing methods that focus only on steady-state performance while neglecting transient smoothness, this invention introduces a smooth control mechanism into event-triggered randomly switched systems. By designing a transition-dependent smooth controller and a modality-dependent stabilizing controller, and optimizing the switching between the two, the double jitter phenomenon caused by "modality switching" and "event triggering" is effectively suppressed, avoiding actuator damage or performance degradation due to drastic changes in control input.

[0045] (2) By optimizing the control law execution method under the event-triggered mechanism, high closed-loop control accuracy is maintained while significantly reducing the online computational load. This invention optimizes the control execution mechanism: at non-triggered moments, the controller does not need to re-solve complex optimization problems, thus greatly saving onboard computing power; at the same time, it uses the real-time system state at the current moment and the pre-stored feedback gain to perform product calculation to generate control commands. This approach not only utilizes the real-time compensation capability of state feedback to ensure rapid suppression of disturbances, but also avoids the high-frequency rolling optimization process through the dynamic triggering mechanism, greatly improving the engineering practicality of the algorithm.

[0046] In summary, this invention addresses the problem of existing semi-Markov system control methods relying too heavily on high-frequency online computation and neglecting smooth transitions in control inputs. It proposes a model predictive control method that combines dynamic event triggering with smooth control. This invention reduces the hardware requirements of the controller and significantly improves the smoothness and reliability of stochastic switching systems during transient processes, providing crucial technical support for the control of complex engineering systems such as UAVs and power systems.

[0047] This invention is applicable to the control of complex engineering systems such as drones and power systems. Attached Figure Description

[0048] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 A flowchart of a semi-Markov smooth model predictive control method for an event-triggered air-to-ground dual-purpose UAV.

[0050] Figure 2 This is a schematic diagram of the ground modes of an air-to-ground dual-purpose unmanned aerial vehicle (UAV) system.

[0051] Figure 3 This is a system state response curve diagram;

[0052] Figure 4 The control input curve corresponding to the designed smooth controller;

[0053] Figure 5 A comparison chart of control input jitter with and without a smoothing controller;

[0054] Figure 6 This is a diagram illustrating the trigger moment. Detailed Implementation

[0055] The specific implementation method of this embodiment, a semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV based on event triggering, includes:

[0056] Traditional event-triggered methods often focus only on system control performance and the number of triggers. However, for air-to-ground UAVs, minimizing control input jitter is a critical issue. This invention applies event-triggered model predictive control to semi-Markov systems. Addressing the widespread external disturbances, it updates the definition of control input smoothness and incorporates it into the infinite-time-domain model predictive control framework. Furthermore, considering the practical needs of limited computational resources, it proposes a new event-triggered strategy to further reduce the number of triggers.

[0057] The event-triggered semi-Markov smooth model predictive control method for air-to-ground dual-use UAVs is implemented as follows:

[0058] Step 1: Determine the UAV system model and initial state, and specify the algorithm parameters:

[0059] A discrete-time unmanned aerial vehicle (UAV) system model incorporating external disturbances and multi-modal switching characteristics is established based on a semi-Markov random process. The initial state and initial modes of the system are determined. Parameters satisfying the dynamic event triggering mechanism are set, and the weight matrix and performance index constants for model predictive control are selected.

[0060] Step 2: Determine if the event has been triggered at the current moment:

[0061] At each sampling moment, the current system state is acquired. The error difference between the current state and the state at the previous trigger moment is calculated. Combined with auxiliary dynamic variables, a judgment is made based on the preset dynamic event triggering conditions. If the triggering conditions are met (i.e., the error exceeds the dynamic threshold) or the current moment is the initial moment, it is determined to be triggered, and the process proceeds to step three to update the control law; if the triggering conditions are not met, it is determined to be not triggered, the optimization solution process is skipped, and the process proceeds directly to step four.

[0062] Step 3: Solve the optimization problem to obtain the feedback control gain and trigger matrix:

[0063] When an event is triggered, an infinite time-domain model predictive control optimization problem, incorporating smoothness and stability constraints, is constructed based on the current system modes and states. Using linear matrix inequalities, this optimization problem is solved online to calculate the positive definite matrix and decision variables that minimize the cost function, thereby obtaining the updated smooth feedback control gain and trigger matrix. This process aims to ensure that the system satisfies recursive feasibility and mean-square input state stability, while limiting the jitter amplitude of the control input.

[0064] Step 4: Based on the current system state, select the control gain to use and calculate the control input:

[0065] Based on the current system operating state (either in a transient state of mode switching or a steady state of mode maintenance), a control gain is selected from the set calculated in step three. Specifically, if the system is in a transitional phase after mode switching or event triggering, a transition-dependent smooth controller is selected to suppress control input jitter; if the system is in a stable phase of modal operation, a mode-dependent stabilizing controller is selected to maintain system performance. Using the selected controller and the current system state, the current control input is calculated.

[0066] Step 5: Apply control inputs to the system, then proceed to Step 2:

[0067] The calculated control inputs are sent to the actuators and applied to the UAV system, driving the system state update to the next moment. The system then operates in a closed loop, returning to step two to perform the event triggering judgment and control loop for the next moment, until the controller design is complete or the mission is finished.

[0068] This implementation addresses the challenges of computational resource constraints and control input jitter faced by semi-Markov transition systems in practical applications. It proposes a dynamic event-triggered model predictive control framework for smooth control and achieves optimized control of the system under disturbed environments by designing a transition-dependent smooth controller and a mode-dependent stabilizing controller. This implementation directly designs the model predictive control law through state observation and a dynamic triggering mechanism, eliminating the need for complex online optimization calculations at every moment, thus achieving efficient control of randomly switching systems. Through this method, smooth, stable, and robust control of semi-Markov transition systems is achieved under limited computational resources, effectively handling issues such as mode-switching jitter, event-triggered update jitter, and the effects of external disturbances.

[0069] Example: Figure 1 As shown, the implementation process of the event-triggered air-to-ground dual-purpose UAV semi-Markov smooth model predictive control method of the present invention is as follows:

[0070] Step 1: For example Figure 2 The ground mode of the air-to-ground dual-purpose UAV is shown. Figure 2 In the coordinate system, I is the inertial coordinate system, B is the body coordinate system, E is the intermediate auxiliary coordinate system used to describe the attitude of the body relative to the ground, h1 is the vertical distance from the origin O to the rotor center o1, h2 is the vertical distance from the origin O to the structural center o2, and θ and ψ represent the pitch angle and yaw angle, respectively.

[0071] The semi-Markov jump system model is constructed as follows:

[0072]

[0073] in, For system status, These are platform speed, pitch angle, acceleration, and pitch angular velocity, respectively. To control the input, For external disturbances, the system matrix

[0074]

[0075] Input matrix

[0076]

[0077] perturbation matrix

[0078]

[0079] in, , Represents the equilibrium point parameters and the initial state of the system. , For system modes, Set an upper bound for the residence time as an auxiliary variable. Weight matrix Input constraints , Sampling period Body mass Gravitational acceleration Moment of inertia Body size Disturb the upper boundary Triggering algorithm-related parameters , Let be the perturbation matrix of system mode 1. Here is the perturbation matrix for system mode 2. Here is the perturbation matrix for system mode 3. For modal dwell time, the system's semi-Markov kernel is:

[0080]

[0081] Step 2: Determine if the event has been triggered at the current moment. or If successful, proceed to step three to solve the optimization problem; otherwise, skip to step four. Definition This refers to the nth switch after the t-th trigger time, where Represents the time of the t-th trigger. ,make , This represents the difference between the current system state and the system state at the time of the most recent trigger. The triggering conditions are as follows:

[0082]

[0083] in , This serves as the trigger matrix for subsequent solutions. As an auxiliary variable,

[0084]

[0085] Step 3: Solve the optimization problem to obtain the feedback control gain and triggering matrix. Define the following function:

[0086]

[0087] in , Let be the positive definite matrix that needs to be solved.

[0088] The optimization problem is as follows: At time t, for a given weight matrix Q, R, There exists a positive definite matrix. sum matrix , making

[0089]

[0090]

[0091] in , , ,

[0092]

[0093] in , , ,

[0094]

[0095] in Control gain Trigger matrix .

[0096] Step 4: Based on the current system status, follow the formula Select the control gain to use and calculate the control input.

[0097]

[0098] Step 5: Apply the control input to the system and proceed to Step 2. Send the control input signal calculated in Step 4 to the UAV actuator system, enabling the system to operate according to the designed control strategy within the current sampling period. Subsequently, at the next sampling moment, acquire the updated system state information and recalculate the error between the current state and the state at the previous trigger moment. Combine this with dynamic auxiliary variables to determine whether the event triggering condition is met again. If the triggering condition is met, proceed to Step 3 to resolve the optimization problem and update the control law; if the triggering condition is not met, use the current control gain and continue executing Step 4 to calculate the control input. Through the above closed-loop iterative process, continuous and stable control of the system under limited computational resources is achieved, while balancing the smoothness of the control input and the system performance requirements.

[0099] This application proposes an event-triggered smooth model predictive control method for air-to-ground dual-purpose unmanned aerial vehicles (UAVs), such as... Figures 3-6 As shown, this application can solve both transition-dependent smooth controllers and modally dependent stabilizing controllers simultaneously using fewer computational resources, reducing control input jitter. This method significantly improves the computational efficiency and operational performance of air-to-ground unmanned aerial vehicles (UAVs) while ensuring system stability and feasibility, and has significant scientific and engineering value for promoting the development of control technology for complex unmanned systems.

Claims

1. A semi-Markov smooth model predictive control method for an event-triggered air-to-ground dual-purpose UAV, characterized in that, The method includes: Step 1: Establish a discrete-time UAV system model based on a semi-Markov random process, incorporating external disturbances and multi-mode switching characteristics; determine the initial state and initial mode of the system; set parameters that satisfy the dynamic event triggering mechanism; and select the weight matrix and performance index constants for model predictive control. Step 2: At each sampling time, obtain the current system state; calculate the difference error between the current state and the state at the previous trigger time, and make a judgment based on the preset dynamic event trigger conditions in combination with auxiliary dynamic variables; if the trigger conditions are met or the current time is the initial time, it is determined to be triggered, and proceed to Step 3 to update the control law; if the trigger conditions are not met, it is determined to be not triggered, skip the optimization solution process, and directly proceed to Step 4. Step 3: When an event is triggered, based on the current system modes and states, construct an infinite time-domain model predictive control optimization problem that includes smoothness constraints and stability constraints; solve the optimization problem online using linear matrix inequality techniques, calculate the positive definite matrix and decision variables that minimize the cost function, and thus obtain the updated smooth feedback control gain and trigger matrix to ensure that the system satisfies recursive feasibility and mean square input state stability, while limiting the jitter amplitude of the control input; Step 4: If the system is in the transition phase after mode switching or event triggering, select a transition-dependent smooth controller to suppress control input jitter; if the system is in the stable phase of modal operation, select a mode-dependent stabilizing controller to maintain system performance; use the selected controller and the current system state to calculate the current control input. Step 5: Send the calculated control input to the actuator to apply to the UAV system, driving the system state to update to the next moment; then the system runs in closed loop, jumps back to step 2, and performs the event trigger judgment and control loop for the next moment until the controller design is completed or the task is completed.

2. The event-triggered semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV according to claim 1, characterized in that: In step one, the unmanned aerial vehicle (UAV) system model is as follows: in, Sampling time, For system status, These are platform speed, pitch angle, acceleration, and pitch angular velocity, respectively. To control the input, External disturbances For system modes, For a set of subsystems, It is a system mode The system matrix, It is a system mode The input matrix, It is a system mode The perturbation matrix; System Matrix Input matrix in, , Indicates the equilibrium point parameters. The sampling period is For the body mass, It is the acceleration due to gravity. For rotational inertia, For the body size, The friction coefficients under different modes.

3. The event-triggered semi-Markov smooth model predictive control method for air-to-ground dual-purpose UAVs according to claim 1, characterized in that: Step one includes: setting the upper bound of the residence time, the weight matrix, and the input constraints. upper boundary of disturbance Triggering algorithm related parameters And the system's semi-Markov kernel.

4. The event-triggered semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV according to claim 1, characterized in that: The triggering condition in step two is: in, The set of natural numbers, This refers to the nth switch after the tth trigger time. Represents the time of the t-th trigger. ,make , This represents the difference between the current system state and the system state at the time of the most recent trigger. , This serves as the trigger matrix for subsequent solutions. As an auxiliary variable, 。 5. The event-triggered semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV according to claim 1, characterized in that: The positive definite matrix in step three is ,right The Lyapunov function is defined as follows: in, .

6. The event-triggered semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV according to claim 5, characterized in that: The optimization problems in step three specifically include: At time t, for a given weight matrix Q, R, There exists a positive definite matrix. sum matrix This makes the cost function: in , , , for Natural numbers within, The upper limit of the dwell time, It is a half-Markov nucleus. For the Lyapunov matrix related to the length of stay, These represent the system's stability performance, smoothing performance, and disturbance suppression performance, respectively. The system matrix is ​​the system mode r. Let r be the input matrix of the system mode r; Let be the perturbation matrix of system mode r; in , , , and These are the invariant set matrices within the subsystem and at the mode switching time, respectively, and both are positive definite. It is a positive definite matrix. Representation matrix The Line number Elements at column, Indicate input constraints The One element; in , for Time controller Control gain , .

7. The event-triggered semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV according to claim 1, characterized in that: In step four, the calculation method for the control input is as follows: Based on the current system state, according to the formula Select the control gain to use and calculate the control input. 。 8. A semi-Markov smooth model predictive control method system for an event-triggered air-to-ground dual-purpose UAV, characterized in that: The system has a program module corresponding to the steps of any one of claims 1-7, and executes the steps in the event-triggered air-to-ground dual-purpose UAV semi-Markov smooth model predictive control method when running.

9. A computer device, characterized in that: It includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, it performs the steps of the event-triggered semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The storage medium is used to store a computer program that executes a semi-Markov smooth model predictive control method for an air-to-ground dual-purpose UAV based on event triggering, as described in any one of claims 1-7.

Images