Intelligent reconstruction control method for water outlet process of cross-medium aircraft based on disturbance observation and neural network

Abstract

The invention discloses an intelligent reconstruction control method for a water outlet process of a cross-medium aircraft based on disturbance observation and a neural network, and belongs to the technical field of aircraft control. The method comprises the following steps: firstly, establishing a unified dynamic model of underwater and air sections; designing a nonsingular terminal sliding mode controller to generate a basic control law, and introducing a finite time disturbance observer to carry out online estimation and compensation on system disturbance to form an enhanced anti-interference control law; meanwhile, a parameter setting device based on a neural network is designed, and air section controller parameters are output in a self-adaptive mode according to the underwater section tail state characteristics; during online execution, an enhanced control law is applied to stabilize the posture of an underwater section, then parameters are set through a neural network, the control law is updated, and finally smooth switching to stable control of an air section is achieved. According to the method, through cooperation of disturbance estimation and intelligent parameter reconstruction, the control precision, the anti-interference capability and the overall robustness of the cross-medium process of the aircraft are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of aircraft control technology, specifically to an intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks. Background Technology

[0002] A cross-medium aircraft is an advanced flight platform capable of operating in both underwater and air environments, with broad application prospects in fields such as resource exploration and emergency rescue. Its emergence from the water is a crucial stage in the aircraft's transition from high-speed underwater motion to flight, involving complex coupling of fluid dynamics and aerodynamic effects, making control extremely challenging.

[0003] Traditional control methods often employ a segmented strategy, that is, designing separate controllers for the underwater segment and the aerial segment, such as... PID While control or sliding mode control are available, these methods have significant limitations. Firstly, underwater and aerial dynamics models differ greatly: underwater dynamics are dominated by fluid viscosity, buoyancy, and cavitation effects, with torque characteristics primarily based on gliding torque; whereas aerial dynamics rely on aerodynamic forces, influenced by dynamic pressure and rudder effects. This mismatch in model parameters leads to poor adaptability of single controller parameters, easily causing attitude oscillations upon exiting the water. Secondly, the exit process faces strong uncertainties such as waves, cavitation drag deviations, and aerodynamic coefficient biases. Traditional controllers lack online disturbance compensation capabilities, making them prone to instability during transitions. Furthermore, the aircraft's mass, angle of attack, and other states change drastically at the moment of exit, and fixed-parameter controllers cannot adaptively adjust, further reducing robustness. While existing technologies attempt to improve performance through model prediction or gain scheduling, they fail to achieve synergy between disturbance estimation and intelligent parameter reconstruction, limiting the smoothness and reliability of cross-medium transitions.

[0004] Therefore, developing a control method that can uniformly model disturbances, estimate disturbances in real time, and intelligently adjust parameters has become an urgent issue for improving the stability of cross-medium aircraft during water exit. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks. By using a unified model, disturbance observation and neural network tuning, the control instability problem of the water exit process of a cross-medium aircraft can be solved.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A smart reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks includes the following steps: Step 1: Establish a unified dynamic model of the underwater and airborne sections of the cross-medium aircraft, including system disturbances. Represent the attitude dynamics and kinematics models in state-space form to define state variables and control variables. The aircraft described by the unified dynamic model is the controlled object. Step 2: Based on the unified dynamics model, design a non-singular terminal sliding mode controller, define the tracking error and sliding surface, design the sliding mode reaching law, and generate the basic control law; Step 3: Design a finite-time disturbance observer, introduce auxiliary variables to estimate the system disturbance online, and output the disturbance estimate; Step 4: Feedforward the disturbance estimate to the basic control law to form an enhanced anti-interference control law, and design a parameter tuner based on a neural network to output adaptive parameters of the air segment sliding surface according to the aircraft state. Step 5: Apply the enhanced anti-interference control law to the controlled object to stabilize the attitude of the aircraft in the underwater phase and simultaneously acquire the end-state characteristics of the underwater phase; Step 6: Input the final state features into the neural network parameter tuner to obtain the adaptive parameters, and update the enhanced anti-interference control law accordingly; Step 7: Apply the updated enhanced anti-interference control law to the controlled object to achieve a stable transition of the aircraft from underwater to air flight and finite-time convergence of attitude.

[0007] Furthermore, in step one: The state variables include roll angle, yaw angle, pitch angle, roll rate, yaw rate, pitch rate, mass, and angle of attack; The control parameters include the rudder deflection angle.

[0008] Furthermore, the sliding surface function of the non-singular terminal sliding mode controller described in step two is designed as a power term including the tracking error, and its form is: ; in, Indicates the sliding surface; The tracking error is the system attitude angle. Its first derivative; p , q To control the gain coefficient and ; All are odd numbers, and satisfy the following conditions: ; It is a symbolic function.

[0009] Furthermore, the sliding mode reaching law is in polynomial form, and its expression is: ; in, The reaching law gain coefficient is greater than zero; , The coefficients are power coefficients and satisfy the following conditions: , ; It is a power function.

[0010] Furthermore, the finite-time perturbation observer described in step three is constructed by introducing a set of auxiliary variables, and its dynamic equation is: ; in, This is the output of the disturbance observer; This is the second derivative of the system tracking error; The known system matrix or vector in the unified dynamics model; To control the quantity; The angular velocity state vector; For matrix The derivative with respect to time; The second derivative of the desired attitude angle; The observer gain coefficient is greater than zero; A scaling factor greater than 1; , , It is a sufficiently small positive number; It is a power function; the disturbance estimate is output by the disturbance observer. Provided.

[0011] Furthermore, the parameter tuner based on the neural network mentioned in step four is a single-layer feedforward network, whose input layer receives the end-state features of the underwater section and whose output layer generates the adaptive parameters of the sliding surface of the aerial section.

[0012] Furthermore, the single-layer feedforward network is trained using an error backpropagation algorithm, and its training sample set contains a multi-level combination of underwater end-state features to establish a nonlinear mapping relationship from the input to the output.

[0013] Furthermore, the terminal state characteristics include the mass, angle of attack, and pitch angle of the aircraft at the end of the underwater phase.

[0014] Another objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks.

[0015] Another objective of this invention is to provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program to enable the electronic device to execute the aforementioned intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks.

[0016] This invention provides an intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks, which significantly improves the control performance of the water exit process of a cross-medium aircraft and has the following significant advantages compared with the prior art: First, a non-singular terminal sliding mode controller based on a unified model design solves the parameter adaptability problem caused by the difference in underwater and aerial dynamics. Through the design of power terms on the sliding surface, the system state is ensured to converge within a finite time. Simulation results show that the attitude angle tracking error drops to zero within 3 seconds, avoiding the switching oscillations of traditional piecewise control.

[0017] Secondly, the finite-time perturbation observer achieves real-time compensation for complex perturbations. The observer can estimate within 0.1 seconds. x shaft and y Axis perturbation, converges within 0.4 seconds. z The shaft disturbance effectively suppresses uncertainties such as waves and aerodynamic deviations, and improves anti-interference capability.

[0018] Furthermore, the neural network parameter tuner enables intelligent reconstruction of controller parameters. Using the final underwater state (e.g., mass, angle of attack) as input, it adaptively outputs sliding surface parameters, significantly improving control accuracy and adapting to different outboard conditions, resulting in a high degree of fit. R >0.995.

[0019] In summary, this invention, through collaborative disturbance estimation and intelligent reconstruction, ensures smooth trajectory, stable rudder deflection, and improved overall robustness, providing reliable control support for cross-medium flight. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the neural network architecture of the present invention; Figure 2 The neural network regression analysis diagram of the present invention includes (a) the regression analysis diagram of the training set, (b) the regression analysis diagram of the validation set, (c) the regression analysis diagram of the test set, and (d) the response analysis diagram of all data. Figure 3 This is a schematic diagram of the three-dimensional trajectory of the cross-medium aircraft exiting water at high speed according to the present invention; Figure 4 This is a schematic diagram of the longitudinal plane motion trajectory of the cross-medium aircraft during the water exit process of the present invention; Figure 5This is a schematic diagram of the attitude angle changes of the cross-medium aircraft during the water exit process of the present invention, wherein (a) is a schematic diagram of pitch angle changes, (b) is a schematic diagram of yaw angle changes, and (c) is a schematic diagram of roll angle changes; Figure 6 This is a schematic diagram of the attitude angular velocity change of the cross-medium aircraft during the water exit process of the present invention, wherein (a) is a schematic diagram of the roll angular velocity change, (b) is a schematic diagram of the yaw angular velocity change, and (c) is a schematic diagram of the pitch angular velocity change. Figure 7 This is a schematic diagram of the estimation results of the disturbance observer during the water exit process of the cross-medium aircraft of the present invention. (a) is a schematic diagram of the underwater x-axis disturbance estimation result, (b) is a schematic diagram of the underwater y-axis disturbance estimation result, (c) is a schematic diagram of the underwater z-axis disturbance estimation result, (d) is a schematic diagram of the air segment x-axis disturbance estimation result, (e) is a schematic diagram of the air segment y-axis disturbance estimation result, and (f) is a schematic diagram of the air segment z-axis disturbance estimation result. Figure 8 This is a schematic diagram of the rudder deflection angle change during the water exit process of the cross-medium aircraft of the present invention, wherein (a) is a schematic diagram of the rudder deflection angle change of the cavitation device, (b) is a schematic diagram of the rudder deflection angle change of the roll channel, (c) is a schematic diagram of the rudder deflection angle change of the deflection channel, and (d) is a schematic diagram of the rudder deflection angle change of the pitch channel. Figure 9 This is a schematic diagram illustrating the mass change of the transmedium aircraft according to the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0022] This invention relates to an intelligent reconfiguration control method for the water exit process of a cross-medium aircraft, particularly suitable for solving problems such as model mutations, unknown disturbances, and control parameter mismatches during high-speed water exit. The core of this method lies in achieving smooth transition control of the aircraft from underwater to air through the synergistic effect of a unified dynamic model, non-singular terminal sliding mode control, a finite-time disturbance observer, and a neural network parameter tuner.

[0023] This embodiment provides an intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks, including the following steps: establishing a unified dynamic model of the underwater and aerial sections of the cross-medium aircraft containing system disturbances; representing the attitude dynamics and kinematics models in state-space form to define state variables and control variables; the aircraft described by the unified dynamic model is the controlled object; based on the unified dynamic model, designing a non-singular terminal sliding mode controller, defining the tracking error and sliding mode surface, designing the sliding mode reaching law, and generating the basic control law; designing a finite-time disturbance observer, estimating the system disturbance online by introducing auxiliary variables, and outputting the disturbance estimate value; The disturbance estimate is fed forward to compensate the basic control law, forming an enhanced anti-interference control law. A neural network-based parameter tuner is designed to output adaptive parameters of the airborne sliding surface based on the aircraft's state. The enhanced anti-interference control law is applied to the controlled object to stabilize the aircraft's attitude in the underwater phase and simultaneously acquire the final state characteristics of the underwater phase. The final state characteristics are input into the neural network parameter tuner to obtain the adaptive parameters, and the enhanced anti-interference control law is updated accordingly. The updated enhanced anti-interference control law is applied to the controlled object to achieve a stable transition of the aircraft from underwater to airborne flight and finite-time attitude convergence. The following is a detailed description with reference to specific embodiments.

[0024] Step 1: Establish a unified dynamic model of the underwater and airborne sections of the cross-medium aircraft, including system disturbances. Represent the attitude dynamics and kinematics models in state-space form to define state and control variables. The aircraft described by this unified dynamic model is the controlled object. The state variables include roll angle, yaw angle, pitch angle, roll rate, yaw rate, pitch rate, mass, and angle of attack; the control variables include rudder deflection.

[0025] To achieve stable control throughout the underwater and aerial shuttle process using a single type of controller, the underwater and aerial dynamic models need to be rewritten into a unified control model. According to flight mechanics, the attitude dynamic model of the transmedium aircraft in the aerial segment is as follows: (1) in, These represent the moments of inertia of the aircraft along its three axes. These are the aircraft's roll angle, yaw angle, and pitch angle, respectively. This indicates the thrust of the swaying nozzle at the tail of the aircraft. This is the vertical distance from the point of thrust application to the longitudinal axis of the aircraft. For the dynamic pressure of the aircraft, The characteristic area of ​​the aircraft, The maximum length of the aircraft These are the yaw moment coefficient and pitch moment coefficient of the aircraft, respectively. This is the distance from the point of thrust application to the spacecraft's center of mass. These are the control rudder deflections for the aircraft's roll, yaw, and pitch channels, respectively.

[0026] According to navigation mechanics, the attitude dynamics model of the underwater section of a cross-medium aircraft is as follows: (2) in, For the aircraft subjected to underwater y Axial direction and z Sliding torque in the axial direction.

[0027] The kinematic models of the aircraft underwater and in the air are as follows: (3) in: These represent the aircraft's roll angle, yaw angle, and pitch angle, respectively.

[0028] Define state variables , and air control and underwater control volume Based on the underwater and aerial dynamics characteristics of the aircraft, the attitude dynamics model (Equation (1) and Equation (2)) and the kinematic model (Equation (3)) are unified into a state-space form (Equation (4)): (4) For the air segment: , , , This represents the system uncertainty caused by deviations in aerodynamic coefficients, as well as external environmental disturbances such as wind disturbances; For the underwater section: , , , This represents the system uncertainty caused by inaccurate gliding force modeling and deviations in the cavitation drag coefficient, as well as external environmental disturbances.

[0029] Step 2: Based on the unified dynamics model, design a non-singular terminal sliding mode controller, define the tracking error and sliding surface, design the sliding mode reaching law, and generate the basic control law.

[0030] Define state The tracking error is According to equation (4), its first derivative is... With the second derivative for: (5) The following section designs a non-singular fast-terminal sliding mode controller for control model (5). The sliding surface function of the non-singular terminal sliding mode controller is designed as a power term including the tracking error, and its form is: (6) in, Indicates the sliding surface; The tracking error is the system attitude angle. Its first derivative; p and q are control gain coefficients and ; All are odd numbers, and satisfy the following conditions: ; It is a symbolic function.

[0031] Differentiating equation (6) with respect to time yields: (7) Design of sliding mode convergence law It takes the form of a polynomial, and its expression is: (8) in, The reaching law gain coefficient is greater than zero; , The coefficients are power coefficients and satisfy the following conditions: , ; It is a power function.

[0032] Let the right side of equation (7) and the right side of equation (8) be equal, while due to the real disturbance Unable to obtain, so please ignore for now. The basic control law is obtained. for: (9) design Lyapunov function for: (1.10) Differentiate equation (10) and substitute equation (7) into it: (11) Under the action of the basic control law (9), the disturbance in the control system This cannot be handled, so a disturbance observer needs to be introduced to estimate and compensate for the disturbance.

[0033] Step 3: Design a finite-time disturbance observer to estimate the system disturbance online by introducing auxiliary variables and output the disturbance estimate. The purpose of the disturbance observer is to monitor the disturbance. To estimate and compensate for the impact of disturbances, the design approach involves introducing a set of auxiliary variables, based on the relationship between these auxiliary variables and the error. The difference between them is used to approximate the disturbance. The specific design process is as follows: A finite-time perturbation observer with a model reference is designed for equation (5). The finite-time perturbation observer is constructed by introducing a set of auxiliary variables, and its dynamic equation is: (12) in, This is the output of the disturbance observer; This is the second derivative of the system tracking error; The known system matrix or vector in the unified dynamics model; To control the quantity; The angular velocity state vector; For matrix The derivative with respect to time; The second derivative of the desired attitude angle; The observer gain coefficient is greater than zero; A scaling factor greater than 1; , , It is a sufficiently small positive number; It is a power function; the disturbance estimate is output by the disturbance observer. Provided.

[0034] Observe the second expression in expression (5) When the output of the disturbance observer and When they are equal, If both are 0, then equation (13) will degenerate into: (13) Observe the first expression of expression (13), when hour: (14) because Therefore, we can conclude that at this time ,because Since it is a known matrix, the external disturbance... An accurate estimate was obtained. In order to achieve... It is necessary to Real-time adjustments are achieved through... State error is introduced in the item. Feedback items If estimated ,but ,at the same time Decrease ,and then Decrease , Decrease, get closer ;when The same principle applies at other times. This method allows for continuous adjustment. Ultimately achieved .

[0035] To analyze the convergence performance of the perturbation observer (Equation (12)), its observation error is defined as: (15) Through derivation, the dynamic characteristics of this observation error satisfy the following differential equation. Equation (12) can be rewritten as: (16) Theoretical analysis shows that the system described by equation (16) is stable in finite time. This means that the disturbance observer designed in this invention can accurately estimate the disturbance within a very short time after the external disturbance changes, providing a prerequisite for real-time compensation by the controller. Therefore, the disturbance observer (12) can converge in a finite time, i.e., when... hour, Considering the disturbance observer's estimation of the disturbance and its compensation to the controller, the designed control law is... Updated to: (17) The control scheme combining the aforementioned finite-time disturbance observer (Equation (12)) with the non-singular terminal sliding mode controller (Equation (9)) ensures that the system tracking error converges within a finite time, and the simulation results below verify this conclusion. This characteristic ensures that the aircraft has a fast response capability and strong robustness when facing model uncertainties and external disturbances.

[0036] Step 4: Feedforward the disturbance estimate to the basic control law to form an enhanced anti-interference control law, and design a neural network-based parameter tuner to output adaptive parameters of the airborne sliding surface based on the aircraft state. The neural network-based parameter tuner is a single-layer feedforward network, whose input layer receives the end-state characteristics of the underwater segment, and whose output layer generates the adaptive parameters of the airborne sliding surface.

[0037] Because the attitude models for the air and underwater segments differ significantly, using a single set of control parameters for both segments would lead to poor parameter adaptability, inadequate control performance, and even divergence in the air segment. Therefore, separate sets of control parameters should be designed for the underwater and air segment controllers. Since the aircraft's state at the end of the underwater segment and its mass characteristics change with different exit angles, the control parameters for the air segment should be capable of adaptive adjustment based on the aircraft's state at the final second of the underwater segment. Therefore, an air flight control parameter tuner based on neural network parameters was designed.

[0038] like Figure 1 As shown, in a single-layer neural network, the input vector Each element is passed through the weight matrix It connects to each neuron. Each neuron combines all weighted inputs and biases Add them together to get its own scalar output. Different Together they form a Network input vector of elements Finally, the network layer outputs a column vector. In summary, a neural network is essentially the construction of a function. By continuously adjusting the weight matrix using the existing training sample set. and bias Improvement function The fitting effect on the parameters in the training sample set. The neural network obtained the optimal weight matrix after training. and optimal bias When using it, when a new input... When the data is fed into the neural network, the neural network model will follow the optimal weight matrix. and optimal bias Output values ​​that match the training results It is worth noting that the training process of a neural network (i.e., a single-layer feedforward network) (i.e., the weight matrix) and bias The adjustment process is performed using an error backpropagation algorithm, which can rely on existing toolkits configured in commercial software, such as... DeepLearningToolbox Training will be conducted, therefore specific and The adjustment process will not be elaborated here. Its training sample set contains multi-level combinations of underwater end-state features to establish a nonlinear mapping relationship from the input to the output.

[0039] For the aerial phase of a cross-medium aircraft, in order to meet the adaptive stability control requirements of different masses, initial angles of attack, and exit angles of water, the final state characteristics (mass, angle of attack, and pitch angle) of the underwater phase model (Equations (2) and (3)) are used as the input to the neural network, i.e. , These represent the mass, angle of attack, and pitch angle at the end of the underwater phase, respectively. The output consists of the sliding surface control parameters that require adaptive adjustment. To reduce the difficulty of neural network fitting, the adaptive control parameters are chosen to be the sliding surface parameters. ,Right now .

[0040] For the training sample set, in order to include as many possible effluent conditions as possible, an effluent quality of 150~300 was selected. kg A total of 27 operating conditions were set up, consisting of initial angles of attack of -2°, 0°, and 2°, and initial pitch angles of 3°, 5°, and 10°, and control parameters were set for each condition. The selected samples form a multi-level combined sample set, which is used by the neural network to construct the mapping function. The specific sample set is shown in Table 1.

[0041] Table 1 Sample Set Table

[0042] Based on the above sample set, 70% was selected as training data, 15% for validation, and 15% for testing. The resulting neural network training results are as follows. Figure 2 As shown. From Figure 2 The neural network regression plot shows that the training set Validation set and test set The values ​​are all greater than This indicates that the mapping results established by the neural network have a high degree of fit and can be used for the output of control parameters in the air section of the subsequent water discharge process simulation.

[0043] After completing the design of the unified dynamic model, non-singular terminal sliding mode controller, finite-time disturbance observer, and neural network parameter tuner, the following intelligent reconfiguration control can be executed online.

[0044] Step 5: Apply the enhanced anti-interference control law to the controlled object to stabilize the underwater attitude of the aircraft and simultaneously acquire the end-of-underwater state characteristics. First, set the initial state of the underwater segment of the aircraft, including the initial attitude angles (roll angle, yaw angle, pitch angle), angular velocity, depth, and initial mass, and simultaneously set the desired attitude angles of the underwater segment (such as pitch angle). Based on the established unified dynamic model (Equation (4)), the underwater controller adopts a non-singular terminal sliding mode control architecture, and its control law is specifically defined by Equation (17). By selecting the underwater segment control parameters (such as sliding surface parameters)p , q Real-time generation of underwater control commands u Control commands u The underwater dynamics model (Equation (2)) and kinematics model (Equation (3)) are applied to achieve underwater attitude stability control. In the last second before emerging from the water (e.g., triggered by a timer or altitude sensor), the end-state characteristics of the underwater segment are synchronously collected, including the aircraft mass, angle of attack, and pitch angle. By combining the unified model with the anti-interference control law, the underwater attitude stability is ensured, and accurate initial state data is provided for the adaptive parameters of the air segment.

[0045] Step Six: Input the final state features into the neural network parameter tuner to obtain the adaptive parameters, and update the enhanced anti-interference control law accordingly. Use the obtained underwater terminal state features (mass, angle of attack, pitch angle) as input. The trained neural network parameter tuner is then input. This network is a single-layer feedforward structure, and its mapping function is... The neural network outputs adaptive parameters of the sliding surface in the air segment based on the input features. p and q The adaptive parameters output by the neural network are then updated into the enhanced anti-interference control law (Equation (17)), replacing the original underwater segment parameters to form a control law optimized for the aerial segment. By intelligently tuning the parameters through the neural network, the parameter mismatch problem caused by the difference between the underwater and aerial models is overcome, and the smooth switching of controller parameters is achieved.

[0046] Step 7: Apply the updated enhanced anti-interference control law to the controlled object to achieve a stable transition of the aircraft from underwater to air flight and finite-time attitude convergence. Apply the updated enhanced anti-interference control law to the air segment dynamics model (Equation (1)) and kinematics model (Equation (3)) and set the desired attitude angle for the air segment. The control law generates rudder deflection commands in real time according to the air segment state. u The attitude of the aircraft is driven to track the desired trajectory. Uncertainties such as aerodynamic disturbances are estimated and compensated in real time by a finite-time disturbance observer (Equation (12)) to ensure that the system state converges in a finite time.

[0047] By reconstructing the control law online and compensating for disturbances, a smooth transition during the water exit process of the cross-medium aircraft and finite-time stable convergence of attitude were achieved, effectively improving anti-interference capability and overall control accuracy.

[0048] In addition, to verify the effectiveness of the intelligent reconfiguration control algorithm proposed in this patent, a simulation of the water exit process control of a cross-medium aircraft was conducted.

[0049] The initial underwater speed of the aircraft was 120. m / sInitial pitch angle 2°, initial yaw angle 0.5°, initial roll angle 5°, initial roll, yaw, and pitch angular velocities all 7° / s Initial depth -10 m The expected underwater pitch angle is 7°, the expected aerial pitch angle is -12°, the expected roll angle and expected yaw angle are 0, and the expected roll, yaw, and pitch velocities are all 0. The initial mass of the aircraft is 250. kg The engine's per second consumption is 2.97. kg / s .

[0050] To verify the effectiveness of the disturbance observer, consider the following external disturbance in the underwater section: (18) External disturbances in the air segment are: (19) The underwater section's gliding moment is deflected by 15%, the cavitation drag coefficient is deflected by 15%, and the airborne aerodynamic coefficient is deflected by 15%. The disturbance observer parameters are: , , , , The control parameters for the shared section between the underwater and aerial sections are as follows: , , , , , , , . Limiting , Rudder deflection angle limit is Control parameters of the underwater sliding surface , The simulation results of the entire process of a cross-medium aircraft exiting water are as follows: Figures 3-9 As shown, the three-dimensional trajectory of the cross-medium vehicle at high speed after exiting the water, the longitudinal plane motion trajectory during the water exit process, the attitude angle change during the water exit process, the attitude angular velocity change during the water exit process, the disturbance observer estimation results during the water exit process, the rudder deflection angle change during the water exit process, and the mass change process are respectively displayed.

[0051] Simulation results show that the cross-medium aircraft exhibits a smooth water exit trajectory, with rapid convergence of attitude angles and angular velocities in both the air and underwater phases, demonstrating high control precision and stable control during the water exit process. Based on the mass, angle of attack, and pitch angle at the end of the underwater phase, the neural network adaptive parameter tuner outputs... , Based on the estimation results from the disturbance observer, the underwater and aerial segments... shaft and Axial disturbances can be mitigated by 0.1 s Convergence, The axial disturbance is 0.4 s Convergence was achieved, verifying the effectiveness of the designed disturbance observer. The change in rudder deflection angle demonstrates that the aircraft can achieve smooth and stable control during its water exit process while meeting rudder deflection angle limits. These results indicate that the present invention effectively improves the aircraft's anti-interference capability and overall stability during cross-medium water exit, ensuring stability and robustness throughout the entire process from water exit to airborne transit.

[0052] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks.

[0053] This embodiment provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor runs the computer program to enable the electronic device to execute the intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks.

[0054] Embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of one or more computer-usable storage media (including, but not limited to, disk storage, etc.) containing computer-usable program code. CD - ROM It takes the form of a computer program product implemented on (such as optical memory, etc.).

[0055] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0056] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0057] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0058] Contents not described in detail in this specification are prior art known to those skilled in the art. It is hereby indicated that the above description is intended to help those skilled in the art understand this invention, but does not limit the scope of protection of this invention. Any equivalent substitutions, modifications, improvements, or simplifications of the above descriptions that do not depart from the essential content of this invention fall within the scope of protection of this invention.

Claims

1. A smart reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks, characterized in that, Includes the following steps: Step 1: Establish a unified dynamic model of the underwater and airborne sections of the cross-medium aircraft, including system disturbances. Represent the attitude dynamics and kinematics models in state-space form to define state variables and control variables. The aircraft described by the unified dynamic model is the controlled object. Step 2: Based on the unified dynamics model, design a non-singular terminal sliding mode controller, define the tracking error and sliding surface, design the sliding mode reaching law, and generate the basic control law; Step 3: Design a finite-time disturbance observer, introduce auxiliary variables to estimate the system disturbance online, and output the disturbance estimate; Step 4: Feedforward the disturbance estimate to the basic control law to form an enhanced anti-interference control law, and design a parameter tuner based on a neural network to output adaptive parameters of the air segment sliding surface according to the aircraft state. Step 5: Apply the enhanced anti-interference control law to the controlled object to stabilize the attitude of the aircraft in the underwater phase and simultaneously acquire the end-state characteristics of the underwater phase; Step 6: Input the final state features into the neural network parameter tuner to obtain the adaptive parameters, and update the enhanced anti-interference control law accordingly; Step 7: Apply the updated enhanced anti-interference control law to the controlled object to achieve a stable transition of the aircraft from underwater to air flight and finite-time convergence of attitude.

2. The intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks according to claim 1, characterized in that, In step one: The state variables include roll angle, yaw angle, pitch angle, roll rate, yaw rate, pitch rate, mass, and angle of attack; The control parameters include the rudder deflection angle.

3. The intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks according to claim 1, characterized in that, The sliding surface function of the non-singular terminal sliding mode controller described in step two is designed as a power term including the tracking error, and its form is: ； in, Indicates the sliding surface; The tracking error is the system attitude angle. Its first derivative; p , q To control the gain coefficient and ; All are odd numbers, and satisfy the following conditions: ; It is a symbolic function.

4. The intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks according to claim 3, characterized in that, The sliding mode reaching law is in polynomial form, and its expression is: ； in, The reaching law gain coefficient is greater than zero; , The coefficients are power coefficients and satisfy the following conditions: , ; It is a power function.

5. The intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks according to claim 1, characterized in that, The finite-time perturbation observer described in step three is constructed by introducing a set of auxiliary variables, and its dynamic equation is: ； in, This is the output of the disturbance observer; This is the second derivative of the system tracking error; The known system matrix or vector in the unified dynamics model; To control the quantity; The angular velocity state vector; For matrix The derivative with respect to time; The second derivative of the desired attitude angle; The observer gain coefficient is greater than zero; A scaling factor greater than 1; , , It is a sufficiently small positive number; It is a power function; the disturbance estimate is output by the disturbance observer. Provided.

6. The intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks according to claim 1, characterized in that, The neural network-based parameter tuner described in step four is a single-layer feedforward network. Its input layer receives the end-state features of the underwater section, and its output layer generates the adaptive parameters of the sliding surface of the aerial section.

7. The intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks according to claim 6, characterized in that, The single-layer feedforward network is trained using an error backpropagation algorithm. Its training sample set contains a multi-level combination of underwater end-state features to establish a nonlinear mapping relationship from the input to the output.

8. A method for intelligent reconfiguration control of the water exit process of a cross-medium aircraft based on disturbance observation and neural networks, as described in claim 1 or 7, characterized in that, The terminal state characteristics include the mass, angle of attack, and pitch angle of the aircraft at the end of the underwater phase.

9. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks, as described in any one of claims 1-8.

10. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory is used to store computer programs and the processor runs the computer programs to enable electronic devices to perform the intelligent reconfiguration control method for the water exit process of a cross-medium aircraft based on disturbance observation and neural networks as described in any one of claims 1-8.

Images