A propeller-driven unmanned vehicle motion control method considering actuator dynamics

Abstract

The application discloses a kind of considering the motion control method of paddle drive unmanned vehicle of actuator dynamics, belong to unmanned vehicle formation control technical field.The existing paddle drive unmanned vehicle is difficult to effectively control in multi-modal switching and high dynamic operation.The geometric relationship between the follower and the follower is established to establish a tracking error model;Construct a finite time preset performance function embedded in overshoot and stable time and other classical control indicators, and perform error transformation through tangent function;Based on the backstepping method, an adaptive control law is designed, a first-order command filter is introduced to suppress calculation expansion, and RBF neural network is used to compensate model uncertainty and external disturbance online.While considering the physical limitations of the actuator, high-precision and strong-robustness motion tracking control of the paddle-driven unmanned vehicle in complex environments is achieved.

Description

Technical Field

[0001] This invention belongs to the field of unmanned vehicle formation control technology, specifically relating to a propeller-driven unmanned vehicle motion control method that takes into account actuator dynamics. Background Technology

[0002] Unmanned aerial vehicles (UAVs) and unmanned vehicles (UAVs) have become crucial equipment for covert reconnaissance, precision detection, and emergency response missions due to their significant advantages of short deployment time and high mobility. In complex scenarios such as urban reconnaissance or disaster search and rescue, single-modal UAVs often struggle to function effectively in varied terrains due to broken and rugged roads and collapsed building walls. To address this challenge, the development of multimodal UAVs with both flight and driving capabilities has become essential. These platforms can switch to flight mode to bypass obstacles in complex terrain and switch to driving mode for stable movement on the ground and even building walls during indoor search and rescue operations. To achieve this multimodal operational capability, propeller-driven UAVs have emerged. These platforms borrow the thrust-adhesion principle from quadcopter UAVs, using propellers mounted on the UAV to generate thrust directed towards the contact surface, thereby obtaining normal pressure and friction from the vertical plane, enabling vertical docking and movement. Typical propeller-driven UAVs typically use a non-powered four-wheeled vehicle chassis and a tandem tilting twin propeller power system. However, in the process of in-depth research on its dynamic modeling and control design, problems such as external disturbances, input saturation, speed constraints and steering failures have become increasingly prominent, placing higher demands on the reliability of the control system.

[0003] Existing control schemes for tracking control of propeller-driven unmanned vehicles (UAVs) often neglect the dynamic characteristics of the actuators themselves. In reality, propeller-driven UAVs rely on motor thrust for longitudinal motion and on servo motors to control the front wheels for lateral motion. The platform's control accuracy is closely related to the dynamics of the motors and servo motors. If the dynamics of the actuators are not considered in the dynamic model, the effectiveness of the control law in practical engineering applications will be directly affected. Furthermore, conventional preset performance control methods often use simple, monotonically decreasing performance functions. While these can constrain error convergence, they lack the ability to quantitatively evaluate control performance, and their performance during transient processes is often unsatisfactory. Current control technologies, while employing algorithms such as backstepping control to address model uncertainties and external disturbances, still have limitations in combining improved control accuracy with quantitative analysis of control performance. Traditional preset performance functions cannot directly embed classic control indices such as settling time and overshoot, resulting in a lack of flexibility in adjusting control time and constraining error fluctuations. Therefore, how to establish a motion control method that considers the dynamic characteristics of actuators and deeply integrates classical control indicators with advanced control algorithms to achieve accurate tracking and performance quantification evaluation of propeller-driven unmanned vehicles in complex environments is an urgent problem to be solved in this field.

[0004] Based on this, the present invention proposes a motion control method for a paddle-driven unmanned vehicle that considers actuator dynamics, in order to solve the problems existing in the prior art. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a motion control method for a propeller-driven unmanned vehicle that considers actuator dynamics, thereby solving the problem that existing propeller-driven unmanned vehicles are difficult to control effectively in multi-modal switching and highly dynamic operations.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A motion control method for a paddle-driven unmanned vehicle that considers actuator dynamics includes the following steps:

[0008] Step S1: Establish a dynamic model of the propeller-driven unmanned vehicle that considers the dynamics of the motor and servo actuator, and establish a tracking error model based on the geometric relationship between the navigator and the follower;

[0009] Step S2: Design a finite-time preset performance function that embeds classical control indices, and use the tangent function to transform the error variables, converting the original bounded error variables into unbounded error variables;

[0010] Based on the backstepping framework, and taking into account the unique thrust adsorption characteristics of propeller-driven unmanned vehicles, the vertical positive pressure generated by the propeller is introduced as a constraint condition into the saturation limit term of the adaptive control law. A first-order command filter is introduced to solve the computational expansion problem, and a neural network is used to compensate for the uncertain disturbances of the system. An adaptive tracking control law is designed to generate the desired motor speed command and the desired servo angle command.

[0011] In a preferred embodiment of the present invention, the process of establishing the dynamic model in step S1 includes:

[0012] Construct the dynamic equations of the moving platform that include the system position vector, velocity vector, and actual output of the actuators as control inputs;

[0013] A first-order inertial dynamic model of the motor is introduced to describe the propeller thrust response, and a dynamic model of the motor is constructed.

[0014] A second-order dynamic model of the servo motor is introduced to describe the front wheel steering process. The second-order differential equation is transformed into a state-space form through auxiliary variables and coupled with the lateral and directional dynamics of the vehicle to establish the dynamic model of the servo motor.

[0015] In a preferred embodiment of the present invention, the process of establishing the tracking error model in step S1 includes:

[0016] Define the line-of-sight distance and line-of-sight angle between the navigator and the follower, and construct the tracking error vector in the vehicle coordinate system;

[0017] The line-of-sight distance error and line-of-sight angle error are defined based on the tracking error vector.

[0018] In a preferred embodiment of the present invention, the finite-time preset performance function with embedded classical control indices described in step S2 for:

[0019] ;

[0020] in, This represents a predefined performance function that is monotonically decreasing. and These represent the upper bound of the error and the final convergence interval of the error, respectively. Indicates finite convergence time. denoted by descent rate, and t represents convergence time.

[0021] In a preferred embodiment of the present invention, step S2, which uses the tangent function to transform the error variable, converts the original bounded error variable into an unbounded error variable as follows:

[0022] ;

[0023] ;

[0024] in, and They represent and The upper realm, and They represent and The lower bound, and They represent and Unbounded variable after transformation by the tangent function This represents the error variable.

[0025] In a preferred embodiment of the present invention, the design steps of the adaptive tracking control law in step S3 include:

[0026] The control system is decomposed into three subsystems: position loop, velocity loop, and control loop.

[0027] The first-level error variable is defined as the transformed position error model, and virtual control quantities are designed layer by layer to stabilize each level of the subsystem.

[0028] In a preferred embodiment of the present invention, the transformed position error model is:

[0029] ;

[0030] ;

[0031] ;

[0032] in, Represents a first-order physical quantity in vector form. Represents the coefficient matrix. Expressing velocity in vector form, Represents the dynamic parameter matrix, Indicates the unknown dynamic part. Indicates external disturbance. Indicates control input, This indicates that the command controls the input. This represents the control input coefficient matrix. The instruction controls the input coefficient matrix. This represents the unknown dynamic part of the model generated by the dynamic modeling of the actuator. Let represent the Gaussian integral function.

[0033] In a preferred embodiment of the present invention, in step S3, the unknown dynamics of the system and external disturbances are estimated online using a neural network compensation term, and the neural network weights are updated adaptively.

[0034] In a preferred embodiment of the present invention, the adaptive tracking control law is composed of three superimposed parts:

[0035] The first part is a proportional term based on error feedback, which is used to provide basic system closed-loop stability;

[0036] The second part is a feedforward term based on the instruction filter, which is used to compensate for the dynamic characteristics of the desired trajectory to improve the response speed;

[0037] The third part is the neural network compensation term, which is used to enhance the robustness of the system under uncertain actuator dynamics.

[0038] In a preferred embodiment of the present invention, the control command finally output by the method directly acts on the motor driver and servo controller of the propeller-driven unmanned vehicle, ensuring that the closed-loop system signal is ultimately consistent and bounded, and that the tracking error meets the preset performance envelope constraint.

[0039] Compared with the prior art, the present invention provides a motion control method for a propeller-driven unmanned vehicle that takes into account actuator dynamics, and has the following beneficial effects:

[0040] 1. This invention solves the problem of frequent overshoot or response lag in traditional control methods caused by neglecting actuator inertia by explicitly including the first-order dynamics of the motor and the second-order dynamics of the servo motor in the modeling. Furthermore, under high-dynamic conditions such as high-speed driving or sharp turns, because the controller pre-calculates the physical limitations of the actuators, the output commands are smoother and more consistent with the hardware execution capabilities, significantly improving the trajectory tracking accuracy of unmanned vehicles in complex terrain.

[0041] 2. This invention establishes a mathematical mapping between classical control indices (overshoot, settling time) and preset performance functions, enabling the performance design of control systems to move beyond tedious parameter trial and error and instead allow for direct setting of control law parameters based on engineering requirements. This significantly enhances the engineering applicability of the algorithm, making the control effect predictable from the design stage.

[0042] 3. This invention, by embedding a preset performance function with classical indices, can finely constrain the convergence trajectory of the error. Compared to traditional monotonically decreasing performance functions, the error trajectory generated by this method, while satisfying rapid convergence, can more effectively suppress system fluctuations when subjected to sudden disturbances, ensuring a smooth transition for the autonomous vehicle during multimodal switching.

[0043] 4. This invention, by incorporating an adaptive control strategy using an RBF neural network, enables the system to learn online and counteract the nonlinear effects caused by load variations, uneven ground friction, and wind disturbances. Even in the absence of precise dynamic parameters, the neural network can maintain high-performance operation of the system through real-time online compensation.

[0044] 5. This invention successfully avoids the "computational bloat" problem commonly found in the control of high-order nonlinear systems by introducing a backstepping method with instruction filtering. This enables complex nonlinear control algorithms to run in real time on low- to medium-performance embedded processors, reducing reliance on onboard computing resources for autonomous vehicles and facilitating large-scale engineering deployment. Attached Figure Description

[0045] Figure 1 The flowchart of the motion control method for a paddle-driven unmanned vehicle that takes into account actuator dynamics is shown in this invention.

[0046] Figure 2 This is an example diagram illustrating the preset performance control of the embedded indicators in this invention.

[0047] Figure 3 This is a trajectory diagram of the propeller-driven unmanned vehicle of the present invention.

[0048] Figure 4 This is a time curve of the follower's line-of-sight distance error in this invention.

[0049] Figure 5This is a graph showing the line-of-sight distance error time curves of followers 1 and 2 in this invention (compared with the Omid Elhaki control method).

[0050] Among them: Figure 4 middle, Figure 4 (a) is a graph showing the line-of-sight distance error for follower 1; Figure 4 (b) is a graph showing the line-of-sight distance error for follower 2.

[0051] exist Figure 5 middle, Figure 5 (a) is a graph showing the line-of-sight distance error for follower 1; Figure 5 (b) is a graph showing the line-of-sight distance error for follower 2. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] As per the instruction manual Figure 1 -Appendix Figure 5 As shown, to address the problem of effective control of existing propeller-driven unmanned vehicles (UAVs) in multi-modal switching and high-dynamic operations, this invention proposes a motion control method for propeller-driven UAVs that considers actuator dynamics, including the following steps:

[0054] Step S1: Establish a dynamic model of the propeller-driven unmanned vehicle that considers the dynamics of the motor and servo actuator, and establish a tracking error model based on the geometric relationship between the navigator and the follower;

[0055] As a novel multimodal mobile platform, propeller-driven unmanned vehicles are powered by tandem tiltable twin propellers, with steering achieved via a front wheel servo motor. To achieve high-precision motion control, it is first necessary to construct a mathematical model that reflects the physical essence of the vehicle.

[0056] Step S1.1: The position of the unmanned vehicle in the ground coordinate system is determined by ( , () indicates that the heading angle of the aircraft system relative to the ground system is and the yaw angle is . The linear velocity of the unmanned vehicle's axis is The dynamic equations of the mobile platform are constructed. These equations include the position vector X. i Velocity vector U i and the actual control input of the actuator output The details are as follows:

[0057] ;

[0058] ;

[0059] ;

[0060] in, Indicates the axle spacing between the front and rear wheels. Indicates the front wheel steering angle. This indicates the linear velocity of the vehicle body at the lower edge of the ground coordinate system. The velocity component, therefore Similarly, ,satisfy , This represents the unknown dynamic components and external disturbances.

[0061] Step S1.2: In practical engineering applications, the motor does not reach the desired speed instantaneously from receiving a voltage command, but rather exhibits electromagnetic and mechanical inertia. Therefore, this embodiment introduces a first-order inertial dynamic model of the motor, constructing the motor's dynamic model as follows:

[0062] ;

[0063] ;

[0064] ;

[0065] in, Indicates the propeller thrust coefficient. Indicates the propeller speed; This indicates the thrust command after considering the dynamic model. This represents the inertial parameter in the dynamic modeling of the motor, which can be identified from the parameters of the thrust test data; This represents an uncertain dynamic model in motor dynamics; and These represent the actual speed of the motor and the desired speed of the motor, respectively.

[0066] Step S1.3: Introduce a second-order dynamic model of the servo motor to describe the front wheel steering process. Transform the second-order differential equation into a state-space form through auxiliary variables and couple it with the lateral and directional dynamics of the vehicle to establish the dynamic model of the servo motor.

[0067] Similar to electric motors, steering servos also exhibit a dynamic response process. The process of a servo driving the front wheels to steer typically exhibits second-order system characteristics, resulting in the following dynamic model of the servo:

[0068] ;

[0069] ;

[0070] in, and These represent the actual servo angle and the desired servo angle, respectively. and These represent the time constant and damping of the second-order system, respectively. These two parameters can also be identified through experimental data parameters. This represents the uncertain disturbance in the steering system. By introducing auxiliary variables, the second-order differential equation of the servo motor is transformed into a state-space form and coupled with the vehicle's lateral angular dynamics to form a complete high-order control model.

[0071] Step S1.4: Based on steps S1.1-S1.4 above, the dynamic model of the propeller-driven unmanned vehicle is obtained as follows:

[0072] ;

[0073] ;

[0074] ;

[0075] in, Represents the control input vector. Indicates the command control input vector. This represents the unknown dynamic components and external disturbances. Represents the dynamic parameter matrix, This represents the control input coefficient matrix. This indicates that the instruction controls the input coefficient matrix. This represents the unknown dynamic part of the model generated by the dynamic modeling of the actuator.

[0076] Step S1.5: To measure the tracking effect, such as... Figure 1 As shown, an error model is established based on the relative geometric relationship between the leader and the follower. Definitions are defined. This indicates the line-of-sight distance between the navigator and the follower. The line-of-sight angle is the angle between the line-of-sight vector and the vehicle body. The angle between them. In the vehicle coordinate system, the tracking error vector is defined as:

[0077] ;

[0078] ;

[0079] ;

[0080] in, Represents longitudinal position error. This represents the lateral position error; , and These represent the navigator's position coordinates and direction, respectively.

[0081] It should be noted that the main objectives of the control strategy designed in this embodiment include the following two parts:

[0082] (1) To keep the desired distance error close to 0, the line-of-sight distance error is defined as:

[0083] ;

[0084] (2) To keep the desired line-of-sight angle error close to 0, the line-of-sight angle error is defined as:

[0085] ;

[0086] The ultimate goal of this method is to design a tracking controller for the transformed dynamic model, ensuring the stability of the closed-loop system and that all signals are eventually uniformly bounded. Furthermore, and It also needs to meet the preset performance requirements.

[0087] Step S2: Design a finite-time preset performance function that embeds classical control indices, and use the tangent function to transform the error variables, converting the original bounded error variables into unbounded error variables;

[0088] Traditional preset performance control typically uses exponentially decreasing monotonically decreasing functions as boundary constraints, but this approach cannot intuitively reflect the overshoot and settling time that are of concern in engineering. This embodiment achieves quantitative evaluation of the control process by embedding classical control indices into the performance function.

[0089] Step S2.1: First, define the finite-time preset performance function. The damping ratio and natural frequency are calculated based on the standard response characteristics of the second-order system. The preset performance function... for:

[0090] ;

[0091] in, This represents a predefined performance function that is monotonically decreasing. and These represent the upper bound of the error and the final convergence interval of the error, respectively. Indicates finite convergence time. Indicates the rate of descent. The larger the value, the steeper the preset performance function. This design ensures that the preset performance function not only has convergence, but its rate of descent and fluctuation characteristics also perfectly match the expected overshoot and settling time, where t represents the convergence time.

[0092] Step S2.2: Construct a classical control envelope boundary curve that shrinks over time. :

[0093] Classical control theory provides the following analytical expression for the step response model of a controlled system:

[0094] ;

[0095] in, , This represents the initial error of the control process; Indicates the damping ratio. This represents the natural frequency. Both can be determined by overshoot. and stable time These two control indicators determine the expression, which is:

[0096] ;

[0097] ;

[0098] ;

[0099] Step S2.3: Determine the correspondence;

[0100] To establish a correspondence between the preset performance function and the classic control index, the preset performance function and the step response curve must satisfy the following: 1) consistent initial error; 2) consistent control accuracy; 3) consistent convergence time; 4) the new preset performance function curve can enclose the step response curve, i.e., a faster initial descent rate. In this way, the preset performance function can meet the requirements for overshoot and settling time. Finally, the following expression can be derived:

[0101] ;

[0102] ;

[0103] ;

[0104] ;

[0105] in, and These represent the upper bound of the error and the final convergence interval of the error, respectively. The first two equations ensure that the initial error and control accuracy are consistent. Indicates the convergence time of the performance function. Ensure consistent convergence time. Indicates the rate of descent. Ensure that the initial descent rate of the preset performance function is faster than the step response curve.

[0106] Step S2.4: Perform error variable transformation based on preset performance;

[0107] To address the nonlinear constraint problem of the control law, this embodiment employs a tangent function for variable transformation. This is because a preset performance function is used to transform the error variable. To impose constraints, the original error variables need to be transformed. This embodiment uses the tangent function for transformation, specifically for the constrained variables. and Bounded errors can be transformed into unbounded errors through the following transformation:

[0108] ;

[0109] ;

[0110] ;

[0111] ;

[0112] in, and They represent and The upper realm, and They represent and The lower bound, and They represent and Unbounded variables after transformation by the tangent function.

[0113] Step S2.5: Solve for the derivative;

[0114] ;

[0115] ;

[0116] ;

[0117] ;

[0118] ;

[0119] ;

[0120] in, and They represent and The total derivative with respect to time. The derivative is found using the chain rule. express right The partial derivatives, Similarly.

[0121] Because the original error variables are constrained, they cannot be directly used in control design. After processing with the tangent function, the bounded original errors are transformed into unconstrained error variables. Therefore, they can be directly rewritten as a vector-form dynamic model, i.e., using... and The dynamic model is derived for first-order variables.

[0122] The dynamic model considering the preset performance can be rewritten as:

[0123] ;

[0124] ;

[0125] ;

[0126] ;

[0127] in, Represents a first-order physical quantity, including and Its derivative can be used to establish a dynamic model. The coefficient matrix is ​​a second-order diagonal matrix, with diagonal elements formed by... and constitute, This represents the unknown dynamic part.

[0128] Based on the anti-stepping framework, and taking into account the unique thrust adsorption characteristics of propeller-driven unmanned vehicles, the vertical positive pressure generated by the propeller is introduced as a constraint condition into the saturation limit term of the adaptive control law. A first-order command filter is introduced to solve the computational expansion problem, and a neural network is used to compensate for the uncertain disturbances of the system. An adaptive tracking control law is designed to generate the desired motor speed command and the desired servo motor angle command.

[0129] For high-order nonlinear systems involving actuator dynamics, this embodiment employs backstepping for control law design. The core idea of ​​backstepping is to decompose the complex system into multiple low-order subsystems and design virtual control quantities layer by layer. However, traditional backstepping suffers from "computational bloat" when calculating the derivatives of the virtual control quantities. Therefore, this embodiment introduces a first-order command filter into the control loop.

[0130] Step S3.1: Define the first-level error variable as the transformed position error model;

[0131] ;

[0132] ;

[0133] ;

[0134] Represents a first-order physical quantity in vector form. Represents the coefficient matrix. Expressing velocity in vector form, Represents the dynamic parameter matrix, Indicates the unknown dynamic part. Indicates external disturbance. Indicates control input, This indicates that the command controls the input. This represents the control input coefficient matrix. The instruction controls the input coefficient matrix. This represents the unknown dynamic part of the model generated by the dynamic modeling of the actuator. Let represent the Gaussian integral function.

[0135] Step S3.2: Control law design;

[0136] Based on backstep control theory, this embodiment sets... and Two dummy variables are used to derive the final control input. To avoid the complexity explosion problem that occurs when applying backstepping control to high-order systems, a first-order filter is used to process the dummy variables. Based on the transformed dynamic equations derived in step S2, the following error variables are set:

[0137] ;

[0138] ;

[0139] ;

[0140] in, This represents the position loop error variable. Represents the velocity loop error variable. This represents the control loop error variable. and They represent and The result after first-order filtering is expressed as:

[0141] ;

[0142] ;

[0143] The unknown dynamics of the system and external disturbances are estimated online using neural network compensation terms, yielding... and The filtering error is:

[0144] ;

[0145] ;

[0146] in, for The filtering error, for The filtering error;

[0147] The control law is as follows:

[0148] Virtual control quantity and Designed as follows:

[0149] ;

[0150] ;

[0151] ;

[0152] ;

[0153] ;

[0154] The command control input is:

[0155] ;

[0156] ;

[0157] ;

[0158] in, and Both represent constant diagonal matrices. and Representing two-dimensional vectors respectively The two elements, Represent the weight matrix, satisfying , Represents radial basis functions. This represents the approximation error of the neural network, and the result converges to a compact set. and They represent The first and second elements, and Design parameters that indicate positive values. Used to estimate the weight matrix , Indicates the estimation error. This represents the upper bound of the estimation error. express initial value, and Both represent positive definite diagonal matrices. and All of these represent design parameters with positive values. Represents a positive definite diagonal matrix. The weight matrix satisfies , This represents the number of hidden layer nodes in a neural network. Represents radial basis functions. This represents the approximation error of the neural network, and the result converges to a compact set. Used to estimate the weight matrix , Indicates external disturbances and The upper bound of the estimate. , , and All of these represent design parameters.

[0159] Through this adaptive mechanism, the neural network can perceive changes in system disturbances in real time and generate compensation terms in the control commands to offset the impact of unmodeled dynamics on tracking accuracy.

[0160] The final control law consists of three parts: the first part is a proportional term based on error feedback, used to ensure the basic stability of the system; the second part is a feedforward term based on a command filter, used to improve the response speed to the desired trajectory; and the third part is a neural network compensation term, used to enhance robustness. The generated control quantity is the motor's desired speed command. and the servo's desired angle command Through this multi-layered nested control structure, the system achieves high-precision tracking of complex trajectories while taking into account the dynamics of the actuators.

[0161] Example: To verify the effectiveness of the above-mentioned propeller-driven unmanned vehicle motion control method considering actuator dynamics of the present invention, this example applies it to a specific unmanned vehicle dynamic model for verification and explanation.

[0162] 1. The dynamic parameters and control parameters of the unmanned vehicle are shown in the table below.

[0163] Table 1: Dynamic Parameters of Propeller-Driven Unmanned Platform

[0164]

[0165] Table 2: Control Parameter Table

[0166]

[0167] Simulation experiments follow Figure 1 Demonstrate the simulation process. Figure 2 An example is given, which is a preset performance function that embeds classical control indices. The numerical simulation of the control algorithm is as follows: Figure 3 and Figure 4 As shown.

[0168] 2. In this embodiment, an example of a preset performance function that embeds classical control indices is given.

[0169] The results are shown in the attached figure. First, the overshoot was set to 0.2, and the settling time was set to 25 steps (S). The initial error was set to 1, and the convergence value was 0.02. Based on the above parameters, the step response curve and theoretical envelope can be calculated. Following the steps proposed in this paper, the preset performance function can be calculated. Final results are attached. Figure 2 As shown, the theoretical envelope includes the step response curve, while The settings ensure that the preset performance function can include initial errors. Ensure that the adjustment time for both is consistent. This allows the preset performance function to have a higher slope at the initial moment, resulting in a faster decrease in the preset performance function. This means that the error variable converges at a higher rate before reaching the convergence time, and the preset performance function has better dynamic performance.

[0170] As attached Figure 3 As shown, under a complex "cross" reference path, the propeller-driven unmanned vehicle exhibits excellent trajectory tracking performance. The red curve represents the reference trajectory, while the green and blue curves represent the actual movement trajectories of the two followers, with a longitudinal span of 4m and a lateral span of approximately 1.2m. A magnified view reveals that even at turns where the path curvature changes drastically, the followers can quickly adjust their course and closely follow the reference path. This demonstrates that the method of this invention, by taking into account actuator dynamics, enables the controller to anticipate and compensate for steering and power system response delays in advance.

[0171] Appendix Figure 4 Demonstrates line-of-sight distance error The curve shows the error over time, with an initial value of 0.45m. The error gradually converges over time and smoothly transitions within the preset performance boundaries. At approximately 2 steps (S), the static error enters the 5% error band, and the final control accuracy reaches 0.0016m. This is based on the line-of-sight error. As can be seen from the expression, when Approaching the preset performance boundary hour, It will approach infinity. The entire convergence process did not exceed the preset envelope boundary, proving that the performance function embedded with classical indices has a very strong ability to constrain error fluctuations.

[0172] Appendix Figure 5 Comparison results between the method of this invention and existing robust adaptive control methods (such as the Omid Elhaki method) are presented. Experimental results show that although both methods eventually achieve error convergence, the method of this invention performs better during the transient process. Specifically, the error curve generated by the method of this invention has a smaller overshoot, and the process to reach steady state is smoother without obvious oscillations. This is because the performance function of this invention includes the damping characteristics of the second-order system, and can automatically adjust the strength of the feedback gain according to the overshoot requirement. In contrast, existing methods, due to the lack of quantified performance constraints, are prone to generating large transient deviations in the initial stage or when disturbances occur.

[0173] In practical hardware deployment scenarios, the method of this invention runs on an airborne embedded computing platform. The motor driver receives a PWM signal (desired speed command) from the controller. The servo receives an angle pulse signal (corresponding to the servo's desired angle command). The sensor system acquires the real-time pose of the autonomous vehicle through IMU and visual positioning. Because a command filter is introduced in the backstepping method, the computational load remains low, and the control frequency can reach over 200Hz, fully meeting the real-time requirements of the propeller-driven autonomous vehicle during high-speed driving and vertical climbing.

[0174] The above embodiments demonstrate that this invention solves the control challenges of propeller-driven unmanned vehicles in multimodal switching and highly dynamic operations by deeply integrating actuator physical characteristics, classical control theory, and advanced adaptive algorithms. The modeling method that considers actuator dynamics eliminates the "gap" between control commands and hardware execution; the preset performance control embedded with classical indicators provides engineers with intuitive performance adjustment tools; and the combination of neural networks and backstepping provides the system with the confidence to cope with complex environments. These technical features collectively ensure that propeller-driven unmanned vehicles can perform stable and precise maneuvering tasks in extreme terrains such as broken roads and vertical walls.

[0175] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for motion control of a paddle-driven unmanned vehicle considering actuator dynamics, characterized in that, Including the following steps: Step S1: Establish a dynamic model of the propeller-driven unmanned vehicle that considers the dynamics of the motor and servo actuator, and establish a tracking error model based on the geometric relationship between the navigator and the follower; Step S2: Design a finite-time preset performance function that embeds classical control indices, and use the tangent function to transform the error variables, converting the original bounded error variables into unbounded error variables; Step S3: Based on the inverse stepping framework, and taking into account the unique thrust adsorption characteristics of the propeller-driven unmanned vehicle, the vertical positive pressure generated by the propeller is introduced as a constraint condition into the saturation limit term of the adaptive control law. A first-order command filter is introduced to solve the computational expansion problem, and a neural network is used to compensate for the uncertain disturbances of the system. An adaptive tracking control law is designed to generate the desired motor speed command and the desired servo angle command. 2.The propeller-driven unmanned vehicle motion control method considering actuator dynamics according to claim 1, wherein, The process of establishing the dynamic model in step S1 includes: Construct the dynamic equations of the moving platform that include the system position vector, velocity vector, and actual output of the actuators as control inputs; A first-order inertial dynamic model of the motor is introduced to describe the propeller thrust response, and a dynamic model of the motor is constructed. A second-order dynamic model of the servo motor is introduced to describe the front wheel steering process. The second-order differential equation is transformed into a state-space form through auxiliary variables and coupled with the lateral and directional dynamics of the vehicle to establish the dynamic model of the servo motor. 3.The propeller-driven unmanned vehicle motion control method considering actuator dynamics of claim 1, wherein, The process of establishing the tracking error model in step S1 includes: Define the line-of-sight distance and line-of-sight angle between the navigator and the follower, and construct the tracking error vector in the vehicle coordinate system; The line-of-sight distance error and line-of-sight angle error are defined based on the tracking error vector.

4. The motion control method for a paddle-driven unmanned vehicle considering actuator dynamics according to claim 1, characterized in that, The finite-time preset performance function with embedded classical control indices described in step S2 for: ； in, This represents a predefined performance function that is monotonically decreasing. and These represent the upper bound of the error and the final convergence interval of the error, respectively. Indicates finite convergence time. denoted by descent rate, and t represents convergence time.

5. The motion control method for a propeller-driven unmanned vehicle considering actuator dynamics according to claim 1, characterized in that, Step S2 describes using the tangent function to transform the error variable, converting the original bounded error variable into an unbounded error variable as follows: ； ； in, and They represent and The upper realm, and They represent and The lower bound, and They represent and Unbounded variable after transformation by the tangent function This represents the error variable.

6. The motion control method for a paddle-driven unmanned vehicle considering actuator dynamics according to claim 1, characterized in that, The design steps of the adaptive tracking control law in step S3 include: The control system is decomposed into three subsystems: position loop, velocity loop, and control loop. The first-level error variable is defined as the transformed position error model, and virtual control quantities are designed layer by layer to stabilize each level of the subsystem.

7. The motion control method for a paddle-driven unmanned vehicle considering actuator dynamics according to claim 6, characterized in that, The transformed position error model is as follows: ； ； ； in, Represents a first-order physical quantity in vector form. Represents the coefficient matrix. Expressing velocity in vector form, Represents the dynamic parameter matrix, Indicates the unknown dynamic part. Indicates external disturbance. Indicates control input, This indicates that the command controls the input. This represents the control input coefficient matrix. The instruction controls the input coefficient matrix. This represents the unknown dynamic part of the model generated by the dynamic modeling of the actuator. Let represent the Gaussian integral function.

8. The motion control method for a paddle-driven unmanned vehicle considering actuator dynamics according to claim 6, characterized in that, In step S3, the unknown dynamics of the system and external disturbances are estimated online using a neural network compensation term, and the neural network weights are updated adaptively.

9. The motion control method for a paddle-driven unmanned vehicle considering actuator dynamics according to claim 1, characterized in that, The adaptive tracking control law is composed of three superimposed parts: The first part is a proportional term based on error feedback, which is used to provide basic system closed-loop stability; The second part is a feedforward term based on the instruction filter, which is used to compensate for the dynamic characteristics of the desired trajectory to improve the response speed; The third part is the neural network compensation term, which is used to enhance the robustness of the system under uncertain actuator dynamics.

10. The propeller-driven unmanned vehicle motion control method considering actuator dynamics according to any one of claims 1-9, characterized in that, The control commands ultimately output by the method directly act on the motor driver and servo controller of the propeller-driven unmanned vehicle, ensuring that the closed-loop system signals are ultimately consistent and bounded, and that the tracking error meets the preset performance envelope constraints.

Images