Optimization method and device for attitude control parameters of carrier rocket based on particle swarm optimization

Abstract

The application provides a launch vehicle attitude control parameter optimization method and device based on particle swarm optimization, and relates to the technical field of launch vehicle control, and comprises the following steps: establishing a closed-loop attitude control system of a launch vehicle; determining a multi-objective optimization function based on the stability margin of the closed-loop attitude control system; establishing a particle swarm based on the control parameters to be optimized, taking the minimization of the multi-objective optimization function as the target, and performing iterative optimization; dynamically adjusting the inertia weight of each particle based on the particle similarity, and constraining the inertia weight based on the individual learning factor and the group learning factor; and determining the optimal control parameter based on the global optimal position after the iterative optimization is completed. The method and device provided by the application can intelligently balance the global exploration and local development capability of the algorithm, improve the search efficiency and optimization accuracy in the parameter space, and accurately design the attitude control parameters of the launch vehicle intelligently.

Description

Technical Field

[0001] This application relates to the field of launch vehicle control technology, and in particular to a method and apparatus for optimizing launch vehicle attitude control parameters based on particle swarm optimization. Background Technology

[0002] Reusable launch vehicles are a crucial means for humanity to achieve large-scale, low-cost space transportation systems, including round-trip flights between Earth and space. However, the design process for launch vehicle control systems is complex due to the numerous adjustable attitude control parameters and constraints. Designers require extensive and iterative adjustments to achieve designs that meet system margin requirements. This is especially true when redesigning rocket configurations or adding functionalities, requiring continuous adjustments to the attitude control parameters as the rocket's dynamics model changes. Designing high-dimensional control parameters not only consumes significant time and effort but also cannot fully guarantee the effectiveness of parameters designed based on experience, directly limiting the design efficiency of launch vehicle attitude control systems.

[0003] Therefore, how to intelligently and accurately design the attitude control parameters of launch vehicles has become a technical problem that the industry urgently needs to solve. Summary of the Invention

[0004] This application provides a method and apparatus for optimizing the attitude control parameters of a launch vehicle based on particle swarm optimization, which is used to solve the technical problem of how to intelligently and accurately design the attitude control parameters of a launch vehicle.

[0005] This application provides a method for optimizing launch vehicle attitude control parameters based on particle swarm optimization, including: A closed-loop attitude control system for a launch vehicle is established; the closed-loop attitude control system includes a coupled dynamic model based on rigid body dynamics equations, swaying dynamics equations and elastic dynamics equations, and an attitude controller with control parameters to be optimized. The multi-objective optimization function is determined based on the stability margin of the closed-loop attitude control system. A particle swarm is established based on the control parameters to be optimized, with the goal of minimizing the multi-objective optimization function. Iterative optimization is performed in the solution space of the control parameters to be optimized. In each iteration, the inertia weight of each particle is dynamically adjusted based on the particle similarity between each particle and the current best particle, and the inertia weight is constrained based on the individual learning factor of each particle and the group learning factor of the particle swarm. The optimal control parameters are determined based on the global optimal position after the iterative optimization is completed.

[0006] In some embodiments, determining the multi-objective optimization function based on the stability margin of the closed-loop attitude control system includes: The current value of each stability margin is mapped to the corresponding reward value of each stability margin through a preset reward function; the preset reward function is used to determine the reward value corresponding to each stability margin based on the closeness between the current value of each stability margin and the target value of each stability margin; the reward value and the closeness are positively correlated. The reward values ​​corresponding to each stability margin are weighted and summed based on the calculation weights corresponding to each stability margin. The multi-objective optimization function is obtained by inverting the weighted summation result.

[0007] In some embodiments, the stability margin includes at least one of low-frequency amplitude margin, high-frequency amplitude margin, rigid body phase margin, sloshing margin, and elastic margin.

[0008] In some embodiments, dynamically adjusting the inertia weight of each particle based on the particle similarity between each particle and the current optimal particle includes: Based on the particle similarity between each particle and the current best particle, a similarity influence factor is determined; the similarity influence factor is inversely correlated with the inertia weight. The iteration process influence factor is determined based on the ratio of the current iteration number to the maximum iteration number; the iteration process influence factor is inversely correlated with the inertia weight. Based on the similarity influence factor and the iteration process influence factor, the inertia weight of each particle is dynamically adjusted.

[0009] In some embodiments, constraining the inertia weight based on the individual learning factors of each particle and the swarm learning factor of the particle group includes: The inertia weight of each particle is constrained based on the average of the individual learning factor of each particle and the group learning factor of the particle swarm.

[0010] In some embodiments, the attitude controller includes a proportional controller, a differential controller, and a higher-order correction network; the higher-order correction network is a higher-order filter composed of multiple second-order notch filters connected in series; the control parameters to be optimized include the proportional gain coefficient, the differential gain coefficient, and the parameters of each second-order notch filter in the higher-order correction network. The method further includes: Based on the importance of each parameter to the system performance, a preset number of parameters are selected as control parameters to be optimized from the parameters of each second-order notch filter in the high-order correction network.

[0011] In some embodiments, the method further includes: Select multiple key flight moments from the flight sequence of the launch vehicle; The optimal control parameters for the flight time between two adjacent critical flight moments are determined by linear interpolation or by keeping them unchanged. Based on the optimal control parameters at each flight time, a set of control parameters covering the flight time sequence is generated.

[0012] This application provides a launch vehicle attitude control parameter optimization device based on particle swarm optimization, comprising: The system construction module is used to establish a closed-loop attitude control system for the launch vehicle; the closed-loop attitude control system includes a coupled dynamic model based on rigid body dynamics equations, swaying dynamics equations and elastic dynamics equations, and an attitude controller with control parameters to be optimized. The function determination module is used to determine a multi-objective optimization function based on the stability margin of the closed-loop attitude control system. An iterative optimization module is used to establish a particle swarm based on the control parameters to be optimized, with the goal of minimizing the multi-objective optimization function, and to perform iterative optimization in the solution space of the control parameters to be optimized; in each iteration of optimization, the inertia weight of each particle is dynamically adjusted based on the particle similarity between each particle and the current best particle, and the inertia weight is constrained based on the individual learning factor of each particle and the group learning factor of the particle swarm. The parameter determination module is used to determine the optimal control parameters based on the global optimal position after the iterative optimization is completed.

[0013] This application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned method for optimizing launch vehicle attitude control parameters based on particle swarm optimization.

[0014] This application provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned method for optimizing launch vehicle attitude control parameters based on particle swarm optimization.

[0015] The method and apparatus for optimizing launch vehicle attitude control parameters based on particle swarm optimization provided in this application employ an improved particle swarm optimization algorithm that dynamically adjusts inertia weights based on particle similarity. This algorithm can efficiently perform global optimization in a high-dimensional parameter space to determine the optimal control parameters of the attitude controller. In each iteration of optimization, the inertia weights of each particle are dynamically adjusted based on the particle similarity between each particle and the current optimal particle. The inertia weights are constrained based on the individual learning factors of each particle and the collective learning factors of the particle swarm. This intelligently balances the algorithm's global exploration and local exploitation capabilities, effectively avoiding the problem of traditional optimization algorithms easily getting trapped in local optima, and improving the search efficiency and optimization accuracy in high-dimensional and complex parameter spaces. Combined with a multi-objective optimization function, it ensures that the optimization results can achieve comprehensive optimization in multiple aspects such as rigid body control, sway suppression, and elastic vibration suppression. This enables intelligent and accurate design of launch vehicle attitude control parameters, improving the efficiency and performance of launch vehicle attitude control system design, and enhancing the stability and robustness of the launch vehicle control system. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is one of the flowcharts illustrating the particle swarm optimization-based method for optimizing launch vehicle attitude control parameters provided in this application.

[0019] Figure 2 This is a schematic diagram of the closed-loop attitude control system for the launch vehicle provided in this application.

[0020] Figure 3 This is a flowchart of the hardware-in-the-loop simulation test of the launch vehicle provided in this application.

[0021] Figure 4 This is the second flowchart of the method for optimizing launch vehicle attitude control parameters based on particle swarm optimization provided in this application.

[0022] Figure 5 This is a schematic diagram of the structure of the launch vehicle attitude control parameter optimization device based on particle swarm optimization provided in this application.

[0023] Figure 6This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first," "second," etc., used in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps, units, or modules is not necessarily limited to those explicitly listed, but may include other steps, units, or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0026] In related technologies, medium and large launch vehicles simultaneously possess complex coupled characteristics of "rigid body-swaying-elasticity," requiring high-order control parameters and a large number of adjustable parameters. Parameter design relies heavily on the engineer's experience, and the design process involves obtaining the closed-loop system transfer function at characteristic second points based on small-perturbation linearized equations. Multidimensional control parameters are then manually adjusted according to phase margin and gain margin requirements, a cumbersome and time-consuming process. Especially during configuration verification and model iteration, control parameters require frequent adjustments, impacting design efficiency, and parameters designed based on the engineer's experience cannot guarantee complete effectiveness.

[0027] In order to address the shortcomings of related technologies, Figure 1 This is one of the flowcharts illustrating the particle swarm optimization-based method for optimizing launch vehicle attitude control parameters provided in this application, such as... Figure 1 As shown, the method includes steps 110, 120, 130 and 140, and is applicable to the attitude control parameter design and optimization scenarios of medium and large reusable liquid launch vehicles.

[0028] Step 110: Establish a closed-loop attitude control system for the launch vehicle; the closed-loop attitude control system includes a coupled dynamic model based on rigid body dynamics equations, swaying dynamics equations and elastic dynamics equations, as well as an attitude controller with control parameters to be optimized.

[0029] Specifically, the execution entity of the launch vehicle attitude control parameter optimization method based on particle swarm optimization provided in this application embodiment is a launch vehicle attitude control parameter optimization device or system. This device can be implemented in software, such as a launch vehicle attitude control parameter optimization program running in a computer system; it can also be implemented in hardware, such as a computer or server that executes the launch vehicle attitude control parameter optimization method based on particle swarm optimization.

[0030] Figure 2 This is a schematic diagram of the closed-loop attitude control system for the launch vehicle provided in this application, as shown below. Figure 2 As shown, the closed-loop attitude control system logically includes a coupled dynamics model as the controlled object and an attitude controller as the control core.

[0031] Coupled dynamics models are used to describe the attitude and motion behavior of launch vehicles during flight with high fidelity. Since medium and large liquid-fueled launch vehicles exhibit multiple characteristics during flight, including rigid body motion, liquid fuel sloshing, and elastic vibration of the rocket structure, and these characteristics are coupled and mutually influential, establishing an accurate coupled dynamics model is fundamental to parameter optimization design. The coupled dynamics model provided in this application includes rigid body dynamics equations, sloshing dynamics equations, and elastic dynamics equations.

[0032] Rigid body dynamics equations describe the six-degree-of-freedom motion (including attitude motion and center-of-mass motion) of a launch vehicle as a rigid body. These equations can be established based on the principles of variable mass dynamics. Sloshing dynamics equations describe the sloshing of the liquid propellant within the rocket body during flight due to factors such as changes in acceleration. These equations can be established using an equivalent simple pendulum model and a mass spring-damper model. Elastic dynamics equations describe the elastic deformation and vibration of the rocket body under the influence of aerodynamic forces and thrust.

[0033] By coupling the above three types of dynamic equations, a high-fidelity dynamic model is formed that can comprehensively reflect the "rigid-sway-elasticity" coupling characteristics of a rocket. Taking a certain type of launch vehicle as an example, considering rigid body, sway, elasticity, and their coupling and interference effects, the following equations can be obtained to describe the coupled dynamic model.

[0034] .

[0035] .

[0036] .

[0037] .

[0038] .

[0039] In the formula, For the angle of attack; The trajectory inclination angle; The pitch angle; For wind attack angle; , and All are coefficients of the elastic equation; and Indicates the order; Indicates the crosslinking coefficient of elastic swaying; Indicates generalized interference force; , and All are coefficients of the swaying equation; The center of mass of the arrow; The equivalent mass centroid of the sloshing liquid portion; These are the coefficients of the rigid body equation; These are, respectively, the aerodynamic damping coefficient, the aerodynamic torque coefficient, and the engine control torque coefficient; These are the force and moment coefficients in the elastic equation; These are the coefficients of the swaying equation; Equivalent disturbance force and torque; For the first Rank elastic generalized coordinates, For the first The frequency of the elastic center vibration, For the first First-order elastic damping ratio; For the first The sloshing displacement of the liquid in each tank The frequency of the shaking. The damping ratio is the sway resistance.

[0040] The attitude controller is the core decision-making component in a closed-loop control system. Its function is to calculate control commands (such as the deflection angle command of the engine nozzle) based on the deviation between the rocket's current attitude and the desired attitude, so as to drive the actuators (such as servo mechanisms) to adjust the direction of the rocket's thrust, thereby eliminating attitude deviation.

[0041] The parameters in the attitude controller can be used as parameters to be optimized. The values ​​of these parameters directly determine the performance of the control system. For example, the attitude controller can be a combination of a proportional-derivative (PD) controller and a correction network. The parameters to be optimized can be selected from the controller's gain coefficient, time constant, coefficients of integral and derivative terms, zero-pole parameters in the correction network, elements in the state feedback matrix, etc.

[0042] Step 120: Determine the multi-objective optimization function based on the stability margin of the closed-loop attitude control system.

[0043] Specifically, stability margin is a key frequency domain indicator in control theory used to measure the robustness of system stability. It represents how much variation in system parameters is allowed to occur while maintaining a stable state. Compared to focusing only on time domain response (such as overshoot and settling time), evaluation methods based on frequency domain stability margin can more profoundly reflect the system's ability to suppress model uncertainties and external disturbances, and are particularly suitable for launch vehicles with complex "rigid-sway-bounce" coupling characteristics.

[0044] The purpose of a multi-objective optimization function is to integrate multiple, even potentially conflicting, stability margin metrics into a single evaluation value. In practical engineering, not all margins are better the larger they are; excessive margins may mean a slow system response or inadequate suppression of certain disturbances. Therefore, it is necessary to construct a comprehensive objective that can balance various performance aspects.

[0045] Step 130: Establish a particle swarm based on the control parameters to be optimized, and perform iterative optimization in the solution space of the control parameters to be optimized with the goal of minimizing the multi-objective optimization function. In each iteration of optimization, dynamically adjust the inertia weight of each particle based on the particle similarity between each particle and the current best particle, and constrain the inertia weight based on the individual learning factor of each particle and the group learning factor of the particle swarm.

[0046] Specifically, the Particle Swarm Optimization (PSO) algorithm can be used for parameter optimization.

[0047] Each particle represents a specific numerical combination of control parameters to be optimized. For example, if the parameters to be optimized include... If there are 1, then the position of each particle is 1. A vector is a dimensional vector, where each component corresponds to a parameter value. The population of all particles is called a particle swarm. The possible range of values ​​for all control parameters to be optimized constitutes the search space of the algorithm, also known as the solution space.

[0048] Each particle has its own position and velocity. In each iteration, the particle updates its velocity and position based on its historical best position and the global best position of the entire particle swarm, thus searching the solution space. Particle swarm optimization (PSO) has advantages such as fast convergence, ease of implementation, and no need for gradient-based optimization, making it suitable for optimizing various complex objective functions.

[0049] In particle swarm optimization (PSO) algorithms, the inertia weight controls the degree to which the speed of the previous iteration affects the current speed, directly impacting the algorithm's search capability. A larger inertia weight is beneficial for global search, while a smaller inertia weight is beneficial for fine-grained local search. In this embodiment, the inertia weight can be dynamically adjusted by calculating particle similarity. Particle similarity can be defined as the proximity between the current position of a particle (i.e., the parameter solution it represents) and the global optimal position of the entire particle swarm. This proximity can be measured in various ways, for example, as a function of the Euclidean distance between two parameter solution vectors in the solution space.

[0050] The specific mechanism for determining inertia weights based on particle similarity can be as follows: When a particle has a high similarity to the particle at the current optimal position (i.e., its position is very close to the global optimal position), it indicates that the particle may already be in a relatively good region. In this case, its inertia weight should be dynamically reduced to slow down its velocity change, thereby enabling a more refined local search within the promising region. Conversely, when a particle has a low similarity to the particle at the current optimal position (i.e., its position is far from the global optimal position), it indicates that the particle may be trapped in a local optimum or in a poor region. In this case, its inertia weight should be dynamically increased to enable it to jump out of the current region at a higher speed and conduct a wider global exploration to find a better solution. Through this dynamic adjustment strategy, the algorithm can maintain a strong global exploration capability in the early stages of optimization, quickly locating promising regions, while performing efficient local fine-grained searches in the later stages of optimization. This improves the convergence speed and optimization accuracy of the algorithm in high-dimensional parameter spaces, effectively avoiding the problem of classical particle swarm optimization algorithms easily getting trapped in local optima.

[0051] Furthermore, embodiments of this application also constrain the inertia weight based on the individual learning factors of each particle and the swarm learning factor of the particle group. The individual learning factor and the swarm learning factor are acceleration constants that control the movement of particles toward their individual optimal positions and the global optimal positions, respectively. By imposing constraints related to these two learning factors on the inertia weight, such as requiring the inertia weight to be greater than a certain threshold related to the learning factors, the convergence of the algorithm can be theoretically better guaranteed, and particle divergence can be avoided during the search process.

[0052] Step 140: Determine the optimal control parameters based on the global optimal position after the iterative optimization is completed.

[0053] Specifically, the iterative optimization process continues until a preset termination condition is met, such as reaching the maximum number of iterations (which can be set according to actual conditions), or the fitness value of the global optimum has not significantly improved for several consecutive generations. After the iteration ends, the algorithm outputs the global optimum position found during the entire search process.

[0054] The global optimal position itself is a vector, and the values ​​of its components constitute the optimal control parameters sought in this application. This set of parameters represents the solution that enables the multi-objective optimization function to reach its optimum (i.e., the best overall performance of various stability margin indices) under the constructed closed-loop attitude control system model.

[0055] The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization provided in this application employs an improved particle swarm optimization algorithm that dynamically adjusts inertia weights based on particle similarity. This algorithm can efficiently perform global optimization in a high-dimensional parameter space to determine the optimal control parameters of the attitude controller. In each iteration of optimization, the inertia weights of each particle are dynamically adjusted based on the particle similarity between each particle and the current optimal particle. The inertia weights are constrained based on the individual learning factors of each particle and the collective learning factors of the particle swarm. This intelligently balances the algorithm's global exploration and local exploitation capabilities, effectively avoiding the problem of traditional optimization algorithms easily getting trapped in local optima, and improving the search efficiency and optimization accuracy in high-dimensional and complex parameter spaces. Combined with a multi-objective optimization function, it ensures that the optimization results can achieve comprehensive optimization in multiple aspects such as rigid body control, sway suppression, and elastic vibration suppression. This enables the intelligent and accurate design of launch vehicle attitude control parameters, improving the efficiency and performance of launch vehicle attitude control system design, and enhancing the stability and robustness of the launch vehicle control system.

[0056] It should be noted that each implementation method of this application can be freely combined, rearranged, or executed individually, and does not need to rely on or depend on a fixed execution order.

[0057] In some embodiments, the attitude controller includes a proportional controller, a differential controller, and a higher-order correction network; the higher-order correction network is a higher-order filter composed of multiple second-order notch filters connected in series; the control parameters to be optimized include the proportional gain coefficient, the differential gain coefficient, and the parameters of each second-order notch filter in the higher-order correction network. The method also includes: Based on the importance of each parameter to the system performance, a preset number of parameters are selected as control parameters to be optimized from the parameters of each second-order notch filter in the high-order correction network.

[0058] Specifically, the attitude controller can adopt a structure based on proportional-derivative (PD) control combined with a frequency correction network. This structure can provide the system with the necessary attitude stiffness (provided by the proportional element) and damping (provided by the derivative element), while flexibly adjusting the system's response characteristics in different frequency bands through the correction network to adapt to the complex "stiff-sway-bounce" coupled dynamic characteristics of the launch vehicle.

[0059] Structurally, an attitude controller can include a proportional controller, a derivative controller, and a higher-order correction network.

[0060] The proportional controller responds to the attitude error itself, while the derivative controller responds to the rate of change of the attitude error (i.e., the attitude angular velocity error). Together, they achieve rapid and stable tracking of the rocket's attitude.

[0061] Proportional (PP) and derivative (D) controllers can be collectively referred to as PD controllers. PD controllers provide fundamental stability and rapid response. Proportional (P) control determines the speed of response to deviations, while derivative (D) control anticipates trends, increases damping, and reduces overshoot and oscillations. It is responsible for stabilizing the overall system and making rapid responses. The control input generated by the PD controller, as the network input of a higher-order correction network, contains elastic vibrations and fluid sloshing signals that are detrimental to the stability of the rocket body. These signals need to be filtered and corrected at several specific modal frequencies using a higher-order correction network.

[0062] The specific form of a higher-order compensation network can be any linear network used by those skilled in the art to adjust the open-loop frequency characteristics of a system. For example, a higher-order compensation network can be a complex network composed of multiple basic compensation units (such as lead, lag, and notch filters) connected in series. In a specific embodiment, the higher-order compensation network is a higher-order filter composed of multiple second-order notch filters connected in series. A second-order notch filter is a special type of band-stop filter with an extremely narrow stopband, which can accurately filter out or significantly attenuate a specific, persistent interference frequency. When the system has multiple specific, harmful mechanical resonant frequencies, each second-order notch filter is precisely tuned to one of the interference frequencies. Connecting them in series constitutes a higher-order compensation network that can simultaneously and accurately suppress elastic vibrations and fluid sloshing at specific frequencies without affecting the normal control performance of the system at other frequencies.

[0063] Accordingly, the control parameters to be optimized may include the proportional gain coefficient, the differential gain coefficient, and the parameters of each second-order notch filter in the higher-order correction network. The parameters of the second-order notch filter may include the notch frequency (defining the center frequency point at which the filter needs to be strongly attenuated), the quality factor (determining the width of the notch bandwidth, i.e., frequency selectivity), and the notch depth (representing the degree of signal attenuation at the notch frequency), etc.

[0064] The attitude controller structure employs a combination of a PD controller and a high-order correction network. The inputs are attitude angle deviation and angular rate. The attitude angle is measured by the inertial navigation system (INS), and the difference between the attitude angle and the programmed command angle yields the attitude angle deviation. The angular rate is obtained by measuring the rate gyroscope. The attitude angle deviation and angular rate are multiplied by the static gain (proportional gain) and dynamic gain (differential gain), respectively, and then passed through the high-order correction network. The sum of these components forms the sway command, which controls the servo mechanism's motion, thereby controlling the engine nozzle's oscillation and the launch vehicle's flight. The static gain, dynamic gain, and parameters in the high-order correction network within the rocket control loop constitute the adjustable attitude control parameters. In liquid-fueled launch vehicles, there are typically dozens of high-order correction parameters. Stability design involves designing appropriate dynamic and static gains and correction parameters to ensure system stability and meet the stability margin requirements of the rocket's closed-loop control system. During rocket flight, the static gain, dynamic gain, and correction parameters in the control loop may change due to aerodynamic and thermal environmental factors, rocket body structural vibration, sudden changes in engine thrust, etc. Therefore, the robustness of the optimized control parameters is reflected in the fact that the control system can better tolerate these external disturbances, enabling the rocket to better resist various extreme external effects.

[0065] The control parameters to be optimized include proportional gain, differential gain, and all parameters in the high-order calibration network. When the calibration network structure is complex, the number of parameters can be very large, for example, reaching dozens. This leads to a sharp increase in the dimensionality of the entire control parameter to be optimized. In a high-dimensional search space, the search efficiency of intelligent optimization algorithms such as particle swarm optimization will decrease significantly, making them more prone to getting trapped in local optima, and convergence to the global optimum requires enormous computational resources and time.

[0066] In high-order calibration networks, not all parameters have an equally important impact on system performance. Some parameters may only have a slight effect on certain minor performance indicators, while others may play a decisive role in the system's critical stability margins (such as sloshing margin and elasticity margin). The importance of each parameter's impact on system performance can be determined based on engineering experience, prior knowledge, or parameter sensitivity analysis results. The mesh parameters can then be ranked according to importance, and a predetermined number of parameters with higher importance can be selected as the control parameters to be optimized. The predetermined number can be flexibly set according to computational resources, optimization time requirements, and optimization accuracy requirements. For example, in a complex calibration network with 20 parameters, analysis may reveal that only 8 parameters truly have a significant impact on the critical margins. In this case, the predetermined number can be set to 8, thereby reducing the optimization dimension from 20 to 8 dimensions.

[0067] The particle swarm optimization-based launch vehicle attitude control parameter optimization method provided in this application uses all parameters in the proportional gain coefficient, differential gain coefficient, and high-order correction network as control parameters to be optimized. This allows the optimization method to not only ensure the stability of rigid body modes but also to selectively reduce gain or compensate phase for specific swaying and elastic modes. This solves the "rigid-swaying-elastic" coupling problem at a deeper level, resulting in a more robust attitude control system with better overall performance. It effectively identifies key parameters that play a decisive role in system performance, thereby significantly reducing the dimensionality of the parameter space to be optimized. This greatly improves the convergence speed and success rate of the subsequent particle swarm optimization algorithm, saving valuable computational resources and design time.

[0068] In some embodiments, the effectiveness of the optimal control parameters is verified by mathematical simulation experiments based on attitude controller-in-the-loop and hardware-in-the-loop semi-physical simulation experiments of key launch vehicle hardware.

[0069] Specifically, the optimal control parameters obtained through particle swarm optimization are theoretical solutions derived under an ideal mathematical model. To ensure that these parameters remain valid and reliable in the real physical world, their validity needs to be verified.

[0070] This application employs a combination of mathematical simulation experiments and hardware-in-the-loop simulation experiments to verify the effectiveness of the optimal control parameters.

[0071] Attitude controller-in-the-loop mathematical simulation involves applying the obtained optimal control parameters to the closed-loop attitude control system and performing a full-system dynamic simulation on a computer; this is a purely digital simulation. By simulating typical flight maneuvers, it verifies whether the control system meets the design requirements in various time and frequency domain indicators.

[0072] Hardware-in-the-loop (HIL) simulation of launch vehicles involves replacing some software models of the closed-loop attitude control system with real physical hardware, thus creating a hardware-in-the-loop simulation. Key hardware refers to components with complex nonlinear and time-delay characteristics that are difficult to model accurately, such as the servo mechanism driving the engine nozzle deflection and the engine nozzle itself. By deploying optimal control parameters in a hardware-in-the-loop simulation environment, the performance of these parameters in the face of non-ideal characteristics of real hardware can be assessed, thereby providing a more comprehensive evaluation of their effectiveness and robustness.

[0073] Only when the optimal control parameters exhibit satisfactory performance in both mathematical simulation and hardware-in-the-loop simulation experiments are they finally confirmed as effective control parameters. If the verification fails, the weights or constraints in the multi-objective optimization function can be readjusted, or the optimization can be performed again, forming an iterative optimization closed-loop process.

[0074] The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization provided in this application verifies the effectiveness of the optimal control parameters by combining mathematical simulation experiments and hardware-in-the-loop simulation experiments, ensuring that the optimal optimized parameters meet the control performance index requirements in both the time domain and the frequency domain.

[0075] In some embodiments, the stability margin includes at least one of low-frequency amplitude margin, high-frequency amplitude margin, rigid body phase margin, sloshing margin, and elastic margin.

[0076] Specifically, low-frequency amplitude margin, high-frequency amplitude margin, rigid body phase margin, sloshing margin, and elastic margin can be selected as stability margins to determine the multi-objective optimization function.

[0077] Low-frequency amplitude margin characterizes the system's ability to track low-frequency commands and suppress low-frequency constant disturbances. High-frequency amplitude margin characterizes the system's ability to suppress high-frequency disturbances such as sensor noise, and its robustness to high-frequency unmodeled dynamics. Rigid body phase margin measures the stability of the rigid body control loop, directly affecting the basic stability of rocket attitude control. Sloshing margin evaluates the control system's suppression effect on liquid sloshing modes, preventing sloshing-control coupling instability. Elastic margin evaluates the control system's suppression effect on rocket body structural vibrations, preventing missile-control coupling instability.

[0078] The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization provided in this application clarifies the key stability margin index used to evaluate the performance of the launch vehicle attitude control system, making the design of multi-objective optimization functions more targeted and comprehensive.

[0079] In some embodiments, determining a multi-objective optimization function based on the stability margin of the closed-loop attitude control system includes: The current value of each stability margin is mapped to the corresponding reward value of each stability margin through a preset reward function. The preset reward function is used to determine the reward value of each stability margin based on the degree of closeness between the current value of each stability margin and the target value of each stability margin. The reward value and the degree of closeness are positively correlated. The reward values ​​corresponding to each stability margin are weighted and summed based on the calculation weights corresponding to each stability margin. The multi-objective optimization function is obtained by inverting the weighted sum.

[0080] Specifically, the attitude control parameter optimization design of medium and large liquid-fueled launch vehicles is a numerical optimization problem with multiple constrained objectives. In the attitude control parameter optimization design of launch vehicles, a reasonable multi-objective optimization function design is crucial for the success of the optimization. It comprehensively considers performance indicators in both the time and frequency domains to ensure that the control system reaches its optimal state in terms of dynamic response and stability. The key to optimization is ensuring that the designed multi-objective optimization function... The value is minimized. The general form of the multi-objective optimization function is as follows: .

[0081] In the formula, For the first A stability margin to be optimized; For the first The calculation weights of the stability margin to be optimized; For the first The target value is the required value of the stability margin indicator to be optimized. For the design of the multi-objective optimization function; The cutoff frequency, This is the minimum constraint value for the cutoff frequency. This represents the maximum constraint value for the cutoff frequency.

[0082] Stability margin measures the stability of a control system under unknown deviations and is a core indicator in the design of launch vehicle attitude controllers. Large liquid-fueled launch vehicles have complex "rigid-sway-elastic" coupling characteristics, and their amplitude-frequency and phase-frequency curves vary in complexity with frequency, requiring the maintenance of stability margins at multiple frequencies. Calculating the stability margin requires first identifying the frequency positions where the margin is considered, and then calculating the corresponding margin value using the amplitude-frequency and phase-frequency characteristics. Considering that the stability requirement of the control system is that the stability margin meets the requirement rather than being excessively large, constraints are added to ensure the approximate range of the margin.

[0083] The key to intelligently optimizing rocket attitude control parameters using an improved particle swarm optimization algorithm lies in designing a reasonable and efficient multi-objective optimization function. In related technologies, the sum of the integrals of the state deviation and the absolute values ​​of the control variables is used as the multi-objective optimization function. However, due to the complex frequency components of large liquid-fueled launch vehicles, the designed correction network has a high order and a large number of parameters. Furthermore, a larger stability margin is not always better; simply adding and inverting the stability margin values ​​is insufficient to obtain a comprehensive multi-objective optimization function. Therefore, a pre-defined reward function can be designed when designing the multi-objective optimization function. This function maps each stability margin value to a reward value near the target value to avoid convergence problems when the target value is too high or too low during the optimization process.

[0084] The pre-defined reward function is used to transform margin values, which have clear physical meaning but differing evaluation criteria, into dimensionless reward values. This reward value intuitively reflects the quality of the current margin value.

[0085] The purpose of the preset reward function is to determine the reward value corresponding to each stability margin based on how close the current value of each stability margin is to its target value. The target value of the stability margin is preset according to engineering design specifications or flight mission requirements; it represents the ideal value or acceptable lower bound of the margin. The reward value is positively correlated with the degree of closeness. This means that the closer the current value of a stability margin is to its target value, or even slightly better, the higher the reward value obtained through the preset reward function mapping; conversely, the greater the difference between the current value and the target value (especially when it is worse than the target value), the lower the reward value, and it may even be negative.

[0086] In one specific embodiment, a preset reward function It can be designed as a piecewise function, whose input is the current value of a certain stability margin and whose output is the corresponding reward value.

[0087] By designing a similar preset reward function that meets the engineering requirements for each stability margin indicator (low-frequency amplitude margin, high-frequency amplitude margin, etc.), the current value of each stability margin can be mapped to the corresponding reward value.

[0088] In this embodiment, the multi-objective optimization function contains metrics such as gain margin and phase margin in the frequency domain, specifically for high and low frequency bands, as follows: .

[0089] In the formula: It is a multi-objective optimization function; These are low-frequency amplitude margin, high-frequency amplitude margin, rigid body phase margin, sloshing margin, and elastic margin, respectively. These are the target values ​​(indicator requirements) for low-frequency amplitude margin, high-frequency amplitude margin, rigid body phase margin, sloshing margin, and elasticity margin, respectively. These are the reward values ​​mapped from low-frequency amplitude margin, high-frequency amplitude margin, rigid body phase margin, sloshing margin, and elasticity margin to the target value. These are the calculation weights for low-frequency amplitude margin, high-frequency amplitude margin, rigid body phase margin, sloshing margin, and elastic margin, respectively. For rigid body frequencies; The cutoff frequency of the rigid body.

[0090] The particle swarm optimization-based method for optimizing launch vehicle attitude control parameters provided in this application clearly and quantitatively transforms complex, multi-dimensional engineering design objectives (meeting various stability margin requirements) into a single-valued, computable multi-objective optimization function. This effectively avoids the one-sidedness of "the larger the margin, the better" in related technology optimizations. Combined with the setting of calculation weights, it allows for flexible adjustment of the importance of various performance indicators according to the specific requirements of different flight missions and different rocket configurations, thereby making the optimization results more in line with actual engineering needs.

[0091] In some embodiments, the inertia weight of each particle is dynamically adjusted based on the particle similarity between each particle and the current best particle, including: Based on the particle similarity between each particle and the current best particle, a similarity influence factor is determined; the similarity influence factor and the inertia weight are inversely correlated. The iteration process influence factor is determined based on the ratio of the current iteration number to the maximum iteration number; the iteration process influence factor is inversely correlated with the inertia weight. The inertia weights of each particle are dynamically adjusted based on the similarity influence factor and the iteration process influence factor.

[0092] Specifically, unlike solid-propellant launch vehicles, liquid-propellant launch vehicle dynamic models are more complex, highlighting the inefficiency of related technologies. There is an urgent need for an intelligent and automated method that can effectively optimize design control parameters in high-dimensional, multi-constraint optimization problems to improve the stability and design efficiency of launch vehicle control systems.

[0093] Particle swarm optimization (PSO) is a swarm intelligence optimization algorithm that obtains optimal parameters by simulating the behavior of groups such as flocks of birds or schools of fish. In optimizing rocket attitude control system parameters, PSO can be used to automatically tune multidimensional parameters in the controller.

[0094] A particle swarm optimization algorithm can be established based on the control parameters to be optimized. The particle swarm optimization algorithm randomly distributes a group of particles in the solution space, where each particle represents a potential solution in the solution space of the control parameters to be optimized. Then, through continuous iteration, the particles update their positions and velocities based on their own historical best solutions and the best solutions in the swarm, and finally find the global optimum, which is the optimal control parameters.

[0095] The formulas for iteratively updating particle velocity and position are as follows: .

[0096] In the formula, and They represent the first time. The th iteration Individual particles Velocity and position in 2D space; , For the first Individual learning factor and group learning factor in each iteration; Inertial weights; and A random number between 0 and 1; For the first The th iteration The historical optimal solution of each particle; For the first The optimal solution in the population during the next iteration.

[0097] The fitness of each particle is calculated, and the control parameters are updated based on the ensemble historical best and the individual historical best, until the convergence condition is met or the maximum number of iterations is reached.

[0098] Because the randomness inherent in the particle swarm optimization algorithm and the initial parameters may lead to some uncertainty in the optimal solution, this application improves the particle swarm optimization algorithm in related technologies by comprehensively considering inertia weights and learning factors to improve its convergence speed and solution accuracy, thereby enhancing the optimization efficiency of the algorithm. In global search, particles need larger inertia weights, while in local search, smaller weights are required. However, fixed weight values ​​cannot effectively balance the contradictions between global and local searches, nor can they enable particles to converge quickly. Classical particle swarm optimization algorithms can solve basic parameter optimization problems, but they suffer from slow convergence speeds, susceptibility to local optima, and the curse of dimensionality when optimizing high-dimensional parameters. Since rocket control systems typically have many parameters to optimize, it is necessary to specifically improve the classical particle swarm optimization algorithm to enhance its applicability in optimizing multi-dimensional parameters of liquid launch vehicle controllers and its global search capability for control parameters.

[0099] Particle velocity updates are highly dependent on inertia weights. During the iteration process, there may be a true global optimum near the current optimal particle. In order to make the velocities of particles close to the current optimal particle smaller, their inertia weights should also be smaller. Therefore, particle similarity can be introduced to design the inertia weight of each particle. Based on the similarity distribution in the particle swarm, the global / local search strategy can be dynamically adjusted.

[0100] The formula for calculating particle similarity is as follows: .

[0101] In the formula: For particles and Particle similarity; For particles and Spatial distance, and These represent the maximum and minimum Euclidean distances between the two particles, respectively. Considering the limited range of controller parameters, these can be selected based on engineering experience. and The specific value to be taken.

[0102] Particle similarity in the above formula Characterizes the first of a set of control parameters to be optimized The particle and the current best particle The degree of proximity between them; spatial distance Characterizes the first of the control parameters to be optimized The particle and the current best particle The Euclidean distance between them, i.e. . For the first The position of each particle. The current optimal particle The location.

[0103] Inertia weight With particle similarity The increase of will gradually decrease, and the calculation formula is as follows: .

[0104] In the formula, This represents the current iteration number of the algorithm. This represents the maximum number of iterations. This represents the minimum value of the inertia weight. This represents the maximum value of the inertia weight. Used as an intermediate substitution variable.

[0105] Based on the formulas for calculating particle similarity and inertia weight, the improved dynamic inertia weight formula can be obtained, namely, the formula for the first dynamic inertia weight. The first particle The expression for the dynamic inertia weight of the dimension parameter is: 。

[0106] In the formula, That is, the improved version of the first The first particle Dynamic inertia weights of dimensional parameters.

[0107] In the above formula, This is a similarity influence factor, and it has an inverse correlation with the inertia weight. It is an influencing factor of the iteration process and is also inversely correlated with the inertia weight.

[0108] The similarity factor and inertia weight are inversely correlated. This means that the higher the similarity between a particle and the current best particle, the smaller its corresponding inertia weight should be; the lower the similarity, the larger the inertia weight should be. The technical principle behind this inverse correlation design is that particles with high similarity are considered to be in a "promising" and excellent region, and their inertia should be reduced to conduct a refined local search; while particles with low similarity may be in a "poor" region or trapped in a local optimum, and their inertia should be increased to encourage them to conduct a large-scale global exploration to escape the predicament.

[0109] The influencing factor of the iteration process is inversely correlated with the inertia weight. This means that as the number of iterations increases, the inertia weight should generally decrease. The technical principle behind this design is that in the early stages of optimization, a larger inertia weight is needed to encourage particles to perform extensive global searches to quickly discover the approximate region that may contain the optimal solution; while in the later stages of optimization, the entire particle swarm may have gathered near the optimal solution, at which point a smaller inertia weight is needed to facilitate fine-grained local searches by the particles, thereby improving the accuracy of the solution.

[0110] The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization provided in this application dynamically correlates the inertial weights of the particle swarm optimization algorithm with particle similarity and iteration process, realizing intelligent and adaptive adjustment of the algorithm's search behavior. Compared with the classic particle swarm algorithm, this method can more effectively search in the high-dimensional and complex solution space of launch vehicle control parameters. While ensuring global exploration capability to avoid premature convergence, it also enhances the local fine search capability near the optimal solution, thereby finding the optimal combination of control parameters that minimizes the multi-objective optimization function faster and more accurately.

[0111] In some embodiments, the inertia weight is constrained based on the individual learning factor of each particle and the swarm learning factor of the particle group, including: The inertia weight of each particle is constrained based on the average of the individual learning factor of each particle and the group learning factor of the particle swarm.

[0112] Specifically, in the particle swarm optimization algorithm, the velocity update of each particle is affected not only by its inertia weight, but also by its individual learning factor. and group learning factor The individual learning factor determines the extent to which a particle learns from its own historical best position, while the group learning factor determines the extent to which a particle learns from the global best position of the entire population.

[0113] To ensure that the particle's trajectory in the search space eventually converges to a stable point or region, rather than diverging infinitely or oscillating continuously, the algorithm's parameters need to satisfy certain mathematical relationships. This application proposes a method for constraining inertia weights, specifically by constraining the inertia weights based on the mean of the individual learning factor and the group learning factor.

[0114] In one specific embodiment, to ensure particle convergence, the learning factor , and The following constraints must be satisfied between them: .

[0115] The particle swarm optimization-based method for optimizing launch vehicle attitude control parameters provided in this application introduces theoretical constraints based on learning factors, building upon the dynamic adjustment of inertia weights. This provides a stronger guarantee for the convergence of the particle swarm optimization algorithm. This design ensures that the dynamic adjustment strategy not only exhibits good optimization performance in practice but is also more robust in theory. It guarantees that when optimizing the parameters of complex systems like launch vehicles, the algorithm can stably and reliably find high-quality solutions without unexpected divergence or failure due to dynamic parameter changes.

[0116] In some embodiments, the effectiveness of the optimal control parameters is verified based on mathematical simulation experiments of the attitude controller-in-the-loop and hardware-in-the-loop semi-physical simulation experiments of the launch vehicle's key hardware, including: The optimal control parameters were applied to the mathematical simulation experiment of the attitude controller in the loop, and the mathematical simulation verification results of the optimal control parameters were obtained. If the mathematical simulation verification results are satisfactory, the optimal control parameters are applied to the hardware-in-the-loop semi-physical simulation test of the launch vehicle's key hardware to obtain the semi-physical simulation verification results of the optimal control parameters. The effectiveness of the optimal control parameters was determined based on the results of hardware-in-the-loop simulation.

[0117] Specifically, ground simulation testing is the primary means of designing and analyzing the rocket attitude stability control system, and a necessary means to preliminarily and comprehensively evaluate the rationality of the attitude stability control scheme and its control parameter design before flight testing. The ground simulation testing in this application embodiment consists of two parts: mathematical simulation testing and hardware-in-the-loop simulation testing.

[0118] In the mathematical simulation test section, after automatically obtaining a set of optimal control parameters using the improved particle swarm optimization algorithm, a stability margin analysis of the rocket attitude control system is first performed based on these optimal control parameters. After meeting the engineering requirements, a controller-in-the-loop (CIL) mathematical simulation test is then conducted. The CIL mathematical simulation test refers to a simulation test that fully includes the coupled dynamics model (the controlled object) and the attitude controller (with configured optimal control parameters). In this simulation test, the simulation program is run to comprehensively obtain the rocket's flight motion state in the time domain to evaluate the flight test results and obtain mathematical simulation verification results. If the mathematical simulation verification results are satisfactory, a hardware-in-the-loop (HIL) simulation test is then conducted to further verify the effectiveness of the parameters.

[0119] In the hardware-in-the-loop (HIL) simulation test section, various nonlinear characteristics exist in the control equipment, such as hysteresis in the servo mechanism. To address the impact of these nonlinear characteristics, tests such as engine nozzle oscillation driven by the servo mechanism are required to analyze whether the optimal control parameters conform to the actual physical characteristics of the HIL system. Due to the inherent limitations of the actual differences between ground and space and ground test conditions, rocket model parameters cannot be obtained completely accurately on the ground. Therefore, a certain deviation range needs to be taken to encompass these differences. This requires the attitude controller design to ensure system stability under various deviation envelopes. Further verifying the rationality and effectiveness of the optimal control parameters through HIL simulation tests is one of the core steps before the rocket's actual flight.

[0120] Hardware-in-the-loop (HIL) simulation of launch vehicles refers to simulation experiments conducted using a simulation computer, interface devices, and real key hardware. The coupled dynamics model of the launch vehicle runs on the simulation computer, while the attitude controller typically runs on another simulation computer or the actual onboard computer. Several key hardware components of the launch vehicle are integrated into this simulation loop. These key components include the launch vehicle's flight controller (as the real hardware for the attitude controller), servo mechanisms, engine nozzles, and inertial navigation system turntable.

[0121] Figure 3 This is a flowchart of the hardware-in-the-loop simulation test of the launch vehicle provided in this application, such as... Figure 3As shown, a hardware-in-the-loop simulation system for a certain type of launch vehicle includes a servo mechanism, engine nozzle, load simulator, flight controller, simulation computer, and a three-axis flight simulation turntable (inertial navigation system turntable). The flight controller loads optimal control parameters to generate control commands (engine yaw commands), which are then executed by the servo mechanism. The servo mechanism drives the actual engine nozzle to yaw. The load simulator applies a physical torque to the output shaft of the servo mechanism. The actual response of the servo mechanism's actions is collected and fed back to the simulation computer in real time. The coupled dynamics model in the simulation computer uses the servo mechanism's response as mechanical input to perform dynamic calculations, obtaining the theoretical flight attitude value of the launch vehicle at the next moment. The simulation computer sends the newly calculated theoretical flight attitude value to the three-axis flight simulation turntable, driving it to rotate to the new attitude. The rotation of the three-axis flight simulation turntable causes the sensors mounted on it to output new attitude measurements to the flight controller.

[0122] The particle swarm optimization-based launch vehicle attitude control parameter optimization method provided in this application combines mathematical simulation experiments and hardware-in-the-loop simulation experiments. This not only ensures the optimality of the final determined control parameters under the ideal model, but more importantly, it ensures their effectiveness and engineering applicability when facing real and complex physical hardware characteristics, greatly reducing the risk of directly applying parameter design to actual flight tests.

[0123] In some embodiments, the method further includes: Select multiple key flight moments in the flight sequence of the launch vehicle; The optimal control parameters for the flight time between two adjacent critical flight times are determined by linear interpolation or by keeping them unchanged. Based on the optimal control parameters at each flight time, a set of control parameters covering the flight time sequence is generated.

[0124] Specifically, in order to intelligently optimize the design of the optimal parameters of the rocket attitude control system throughout the entire flight sequence, it is necessary to optimize the control system parameters for each time segment based on the flight sequence provided by the overall professional team.

[0125] Critical flight moments refer to the points in the entire flight sequence where the rocket's dynamic characteristics or external environment undergo significant changes. For example, critical flight moments may include flight event points, extreme points in the aerodynamic environment, points where system characteristics change significantly, and equally spaced sampling points.

[0126] Between two adjacent critical flight times, other flight times are included. The optimal control parameters for the two adjacent critical flight times can be linearly interpolated or kept unchanged to determine the optimal control parameters for the flight times between the critical flight times. Finally, the optimal control parameters for each flight time are compiled to generate a control parameter set covering the entire flight time series.

[0127] In the engineering practice of launch vehicles, flight timing is usually counted in seconds. Therefore, critical flight moments can be identified as critical characteristic seconds.

[0128] For example, let's take the first-stage flight control of a rocket as an example to illustrate the optimization design of attitude control parameters. From the moment the rocket launches to the moment of first- and second-stage separation, several key characteristic seconds can be selected for segmented optimization design. The optimal control parameters within each segment are optimized segment by segment. The optimal control parameters between two adjacent characteristic seconds can be determined according to the specific flight conditions; that is, they can be designed to be consistent with those at the key characteristic seconds or obtained through linear interpolation. The optimization design approach for control parameters in the second-stage flight segment is consistent with that of the first stage. In this way, the parameters of the rocket's first and second-stage attitude control system throughout the entire flight sequence can be intelligently and automatically optimized based on an improved particle swarm optimization algorithm. The optimal attitude control parameters of the entire rocket system intelligently designed through the above segmented optimization steps, after verifying their effectiveness through attitude control system simulation experiments and hardware-in-the-loop simulation experiments, can be quickly reused in the attitude control system of similar reusable liquid rockets to improve design and testing efficiency.

[0129] The particle swarm optimization-based launch vehicle attitude control parameter optimization method provided in this application determines the optimal control parameters for the flight time between two adjacent critical flight moments by linearly interpolating or keeping the optimal control parameters of two adjacent critical flight moments unchanged. This ensures the optimality of control performance at each critical flight moment and the smooth transition of control parameters throughout the flight process, thereby enabling the launch vehicle to always operate in a near-optimal state, greatly improving its attitude control accuracy, stability, and robustness in complex and time-varying environments.

[0130] Figure 4 This is the second flowchart of the particle swarm optimization-based method for optimizing launch vehicle attitude control parameters provided in this application. Figure 4 As shown in the embodiments of this application, the method provided is applicable to the intelligent optimization design of parameters for the high-dimensional attitude control system of medium and large liquid launch vehicles.

[0131] Taking a certain type of liquid-fueled launch vehicle as an example, the improved particle swarm optimization algorithm can automatically obtain the optimal attitude control parameters of the rocket's three attitude channels by selecting key feature seconds within the flight time sequence for intelligent optimization. The effectiveness of the optimal attitude control parameters designed by intelligent optimization is verified by adopting a cyclic iterative optimization scheme of six-degree-of-freedom mathematical simulation test in the controller-in-the-loop and semi-physical simulation test in the key hardware-in-the-loop.

[0132] The control parameter optimization process is as follows: by applying the three attitude channels to the designed multi-objective optimization function and the frequency domain constraints of the control system, a set of optimal control parameters at specific characteristic seconds can be obtained intelligently and automatically. Then, the optimal attitude control parameters for other characteristic seconds are designed sequentially. After integrating all characteristic seconds, all optimal parameters for the entire flight sequence are designed, thereby improving the stability and robustness of the rocket attitude control system and increasing the design efficiency of the designer in attitude control system and control parameters.

[0133] The attitude control system was configured with three analytical states: rated, upper limit deviation, and lower limit deviation. Based on these, several key feature points were selected within the flight timeline to further analyze the stability margins of rigid body, elastic, and fluid sloshing, taking into account factors such as the servo mechanism characteristic deviation envelope and elastic modal parameter deviations.

[0134] The apparatus provided in the embodiments of this application is described below. The apparatus described below can be referred to in correspondence with the method described above.

[0135] Figure 5 This is a schematic diagram of the structure of the launch vehicle attitude control parameter optimization device based on particle swarm optimization provided in this application, as shown below. Figure 5 As shown, the device includes: System building module 510 is used to establish a closed-loop attitude control system for the launch vehicle. The closed-loop attitude control system includes a coupled dynamic model based on rigid body dynamics equations, swaying dynamics equations and elastic dynamics equations, as well as an attitude controller with control parameters to be optimized. The function determination module 520 is used to determine the multi-objective optimization function based on the stability margin of the closed-loop attitude control system. The iterative optimization module 530 is used to establish a particle swarm based on the control parameters to be optimized, with the goal of minimizing the multi-objective optimization function, and to perform iterative optimization in the solution space of the control parameters to be optimized. In each iteration of optimization, the inertia weight of each particle is dynamically adjusted based on the particle similarity between each particle and the current best particle, and the inertia weight is constrained based on the individual learning factor of each particle and the group learning factor of the particle swarm. The parameter determination module 540 is used to determine the optimal control parameters based on the global optimal position after the iterative optimization is completed.

[0136] The launch vehicle attitude control parameter optimization device provided in this application adopts an improved particle swarm optimization algorithm that dynamically adjusts inertia weights based on particle similarity. This algorithm can efficiently perform global optimization in a high-dimensional parameter space to determine the optimal control parameters of the attitude controller. In each iteration of optimization, the inertia weights of each particle are dynamically adjusted based on the particle similarity between each particle and the current optimal particle. The inertia weights are constrained based on the individual learning factors of each particle and the collective learning factors of the particle swarm. This intelligently balances the global exploration and local development capabilities of the algorithm, effectively avoiding the problem of traditional optimization algorithms getting trapped in local optima, and improving the search efficiency and optimization accuracy in high-dimensional and complex parameter spaces. Combined with a multi-objective optimization function, it ensures that the optimization results can achieve comprehensive optimization in multiple aspects such as rigid body control, sway suppression, and elastic vibration suppression. This realizes the intelligent and accurate design of launch vehicle attitude control parameters, improves the efficiency and performance of launch vehicle attitude control system design, and enhances the stability and robustness of the launch vehicle control system.

[0137] Figure 6 This is a schematic diagram of the structure of the electronic device provided in this application, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communications bus 640, wherein the processor, communications interface, and memory communicate with each other via the communications bus. The processor can invoke logical commands stored in the memory to execute the methods described in the above embodiments, for example: A closed-loop attitude control system for the launch vehicle is established. This system includes a coupled dynamic model based on rigid body dynamics, swaying dynamics, and elastic dynamics equations, as well as an attitude controller with parameters to be optimized. A multi-objective optimization function is determined based on the stability margin of the closed-loop attitude control system. A particle swarm is established based on the parameters to be optimized, and iterative optimization is performed in the solution space of the parameters to be optimized, with the goal of minimizing the multi-objective optimization function. In each iteration, the inertial weights of each particle are dynamically adjusted based on their similarity to the current best particle, and the inertial weights are constrained based on the individual learning factors of each particle and the collective learning factor of the particle swarm. The optimal control parameters are determined based on the global optimal position after the iterative optimization.

[0138] Furthermore, when the logical commands in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several commands to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0139] The processor in the electronic device provided in this application embodiment can call logical instructions in the memory to implement the above method. Its specific implementation method is the same as the aforementioned method implementation method and can achieve the same beneficial effects, which will not be repeated here.

[0140] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the methods provided in the above embodiments.

[0141] The specific implementation method is the same as the aforementioned method implementation method and can achieve the same beneficial effects, so it will not be repeated here.

[0142] This application provides a computer program product, including a computer program that, when executed by a processor, implements the method described above.

[0143] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0144] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for optimizing the attitude control parameters of a launch vehicle based on particle swarm optimization, characterized in that, include: A closed-loop attitude control system for a launch vehicle is established; the closed-loop attitude control system includes a coupled dynamic model based on rigid body dynamics equations, swaying dynamics equations and elastic dynamics equations, and an attitude controller with control parameters to be optimized. The multi-objective optimization function is determined based on the stability margin of the closed-loop attitude control system. A particle swarm is established based on the control parameters to be optimized, with the goal of minimizing the multi-objective optimization function. Iterative optimization is performed in the solution space of the control parameters to be optimized. In each iteration, the inertia weight of each particle is dynamically adjusted based on the particle similarity between each particle and the current best particle, and the inertia weight is constrained based on the individual learning factor of each particle and the group learning factor of the particle swarm. The optimal control parameters are determined based on the global optimal position after the iterative optimization is completed.

2. The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization according to claim 1, characterized in that, The determination of the multi-objective optimization function based on the stability margin of the closed-loop attitude control system includes: The current value of each stability margin is mapped to the corresponding reward value of each stability margin through a preset reward function; the preset reward function is used to determine the reward value corresponding to each stability margin based on the closeness between the current value of each stability margin and the target value of each stability margin; the reward value and the closeness are positively correlated. The reward values ​​corresponding to each stability margin are weighted and summed based on the calculation weights corresponding to each stability margin. The multi-objective optimization function is obtained by inverting the weighted summation result.

3. The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization according to claim 1, characterized in that, The stability margin includes at least one of low-frequency amplitude margin, high-frequency amplitude margin, rigid body phase margin, sloshing margin, and elastic margin.

4. The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization according to claim 1, characterized in that, The dynamic adjustment of the inertia weight of each particle based on the particle similarity between each particle and the current best particle includes: Based on the particle similarity between each particle and the current best particle, a similarity influence factor is determined; the similarity influence factor is inversely correlated with the inertia weight. The iteration process influence factor is determined based on the ratio of the current iteration number to the maximum iteration number; the iteration process influence factor is inversely correlated with the inertia weight. Based on the similarity influence factor and the iteration process influence factor, the inertia weight of each particle is dynamically adjusted.

5. The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization according to claim 1, characterized in that, The constraint on the inertia weight based on the individual learning factor of each particle and the swarm learning factor of the particle group includes: The inertia weight of each particle is constrained based on the average of the individual learning factor of each particle and the group learning factor of the particle swarm.

6. The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization according to claim 1, characterized in that, The attitude controller includes a proportional controller, a differential controller, and a higher-order correction network; the higher-order correction network is a higher-order filter composed of multiple second-order notch filters connected in series; the control parameters to be optimized include the proportional gain coefficient, the differential gain coefficient, and the parameters of each second-order notch filter in the higher-order correction network. The method further includes: Based on the importance of each parameter to the system performance, a preset number of parameters are selected as control parameters to be optimized from the parameters of each second-order notch filter in the high-order correction network.

7. The method for optimizing launch vehicle attitude control parameters based on particle swarm optimization according to any one of claims 1 to 6, characterized in that, The method further includes: Select multiple key flight moments from the flight sequence of the launch vehicle; The optimal control parameters for the flight time between two adjacent critical flight moments are determined by linear interpolation or by keeping them unchanged. Based on the optimal control parameters at each flight time, a set of control parameters covering the flight time sequence is generated.

8. A device for optimizing launch vehicle attitude control parameters based on particle swarm optimization, characterized in that, include: The system construction module is used to establish a closed-loop attitude control system for the launch vehicle; the closed-loop attitude control system includes a coupled dynamic model based on rigid body dynamics equations, swaying dynamics equations and elastic dynamics equations, and an attitude controller with control parameters to be optimized. The function determination module is used to determine a multi-objective optimization function based on the stability margin of the closed-loop attitude control system. An iterative optimization module is used to establish a particle swarm based on the control parameters to be optimized, with the goal of minimizing the multi-objective optimization function, and to perform iterative optimization in the solution space of the control parameters to be optimized; in each iteration of optimization, the inertia weight of each particle is dynamically adjusted based on the particle similarity between each particle and the current best particle, and the inertia weight is constrained based on the individual learning factor of each particle and the group learning factor of the particle swarm. The parameter determination module is used to determine the optimal control parameters based on the global optimal position after the iterative optimization is completed.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for optimizing the attitude control parameters of a launch vehicle based on particle swarm optimization as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for optimizing the attitude control parameters of a launch vehicle based on particle swarm optimization as described in any one of claims 1 to 7.

Images