Marine rocket recovery platform attitude stability control method based on particle swarm optimization

Abstract

The invention discloses an offshore rocket recovery platform attitude stability control method based on particle swarm optimization, and the method comprises the following steps: S1, collecting and preprocessing multi-source attitude data, and generating a time series data set; s2, constructing a fuzzy PID controller, and setting nine three-axis control parameters; s3, performing global optimization by adopting an improved particle swarm algorithm, and outputting an initial parameter solution; s4, introducing a zebra optimization algorithm to carry out local refined optimization; s5, constructing a collaborative optimization architecture, and fusing and outputting optimal control parameters; s6, the optimal parameters are input into a controller, and a servo system is driven to adjust the three-axis attitude; s7, monitoring disturbance amplitude, dynamically triggering a zebra strategy and feeding back the zebra strategy to the particle swarm; and S8, evaluating the control performance, and feeding back iterative optimization if the control performance does not reach the standard. According to the method, the improved particle swarm optimization algorithm and the zebra optimization algorithm are fused, so that self-adaptive optimization and accurate attitude stable control of attitude control parameters of the offshore rocket recovery platform are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent control and aerospace recovery technology, and particularly relates to a sea rocket recovery platform attitude stabilization control method based on particle swarm optimization. BACKGROUND

[0002] With the rapid development of space transportation systems, the development and engineering application of reusable launch vehicles is becoming an important development direction of global aerospace technology. In the recoverable rocket system, the return of the first stage rocket to the recovery platform after completing the boost task is a key link to improve the reuse rate of the rocket and reduce the launch cost. In order to realize safe and precise recovery landing control, it is necessary to ensure that the sea recovery platform has good attitude stability in complex marine environment, so as to suppress the influence of external wind, wave, current, structure coupling disturbance and other factors on the platform motion state, and ensure that the recovery platform can provide a stable and controllable landing base.

[0003] At present, the attitude control system of the sea rocket recovery platform mostly uses traditional PID controller for attitude adjustment, which realizes closed-loop control of the pitch angle, roll angle and yaw angle of the platform by setting three parameters of proportion, integral and differential. Although the PID control structure is simple and easy to implement, when facing the frequently changing nonlinear disturbance in the sea dynamic environment, its control performance often shows slow response, increased jitter, fixed parameters and other problems. In order to improve the adaptability and robustness of the control system, some studies introduce fuzzy control method, that is, fuzzy rules are used to perform fuzzy reasoning and adjustment on attitude error, so as to weaken the influence of environmental uncertainty. However, the nine control parameters (proportion gain, integral gain and differential gain in each channel) of the fuzzy PID controller have a significant influence on the attitude control performance, and how to quickly and accurately optimize these control parameters becomes a core challenge in the current application of fuzzy control.

[0004] In solving the above problems, swarm intelligence optimization algorithm is widely used in the optimization task of fuzzy control parameters due to its good global search ability and parameter self-adjusting mechanism. Among them, particle swarm optimization algorithm (PSO) is widely used in control system parameter setting due to its simple structure, fast convergence speed and low implementation cost. However, the standard particle swarm optimization algorithm often faces problems such as premature convergence, falling into local optimum and degeneration of search ability when facing complex nonlinear systems, especially control problems with frequent disturbances and prominent dynamic characteristics, which limits its application effect in high-precision control scenarios.

[0005] To improve the practicability of particle swarm optimization algorithm in complex dynamic environment, researchers try to improve it by parameter self-adaption, mutation mechanism introduction, hybrid optimization strategy and other methods. For example, the introduction of inertia weight dynamic adjustment mechanism realizes the smooth transition from global exploration to local convergence, or the use of adaptive learning factor enhances the direction adjustment ability of the algorithm. To some extent, these methods improve the search stability and optimal solution quality, but they still lack deep perception ability of disturbance state and cannot dynamically adjust the optimization strategy according to the change of environmental disturbance, which is not adaptive enough.

[0006] In addition, some studies introduce other heuristic algorithms and particle swarm optimization algorithm for joint optimization, such as genetic algorithm, ant colony algorithm, etc., in order to overcome the limitations of single strategy through the synergistic optimization of multiple algorithms. However, most of these methods are simple in structure and lack of master-slave collaborative mechanism and feedback regulation structure, which cannot fully utilize the optimization advantages of each algorithm, resulting in poor fusion effect and even introducing redundant calculation, affecting control efficiency and real-time performance. Especially in high dynamic, strong disturbance and high safety systems such as offshore rocket recovery platform, the existing methods lack an optimization framework that can simultaneously consider control accuracy, disturbance adaptability and computational efficiency.

[0007] On the other hand, in the existing optimization process, the coupling evolution relationship between control parameters and platform actual attitude response is often ignored, and the linkage feedback between controller design and platform physical behavior cannot be realized. Traditional methods rely more on a single fitness index to evaluate the pros and cons of control parameters statically, ignoring the synergistic optimization of multi-dimensional performance indicators such as dynamic stability of platform attitude response, disturbance recovery speed and control energy consumption. In addition, there is a lack of backtracking mechanism and similarity evaluation means between particle evolution path and platform response behavior, which also limits the optimization algorithm's ability to deeply mine and reuse stable control structures.

[0008] In summary, the existing technology has the following defects when facing the attitude stabilization control of offshore recovery platform: first, there is a lack of intelligent control method that can adaptively adjust the optimization strategy under dynamic disturbance conditions; second, the control parameter optimization process is decoupled from the actual attitude response, and a performance-driven feedback loop mechanism cannot be established; third, the multi-algorithm fusion structure is mostly in a weak cooperative mode, and a master-slave linkage and complementary enhancement cooperative optimization mechanism has not been formed; fourth, there is a lack of structural correlation between particle search path and platform control behavior, which affects the parameter learning efficiency and model generalization ability. These technical bottlenecks seriously restrict the engineering application of fuzzy PID control in high dynamic attitude control systems, and an intelligent attitude control method that integrates deep optimization structure, disturbance perception mechanism and feedback-driven strategy is urgently needed.

[0009] Therefore, how to provide an offshore rocket recovery platform attitude stabilization control method based on particle swarm optimization is a problem that needs to be solved by those skilled in the art. SUMMARY

[0010] An object of the present application is to propose a particle swarm optimization-based offshore rocket recovery platform attitude stabilization control method. The present application combines fuzzy PID control technology, particle swarm optimization algorithm and zebra optimization algorithm, constructs a master-slave collaborative optimization architecture for accurate and real-time control of the pitch angle, roll angle and yaw angle of the recovery platform, and realizes efficient setting of control parameters and continuous optimization of attitude response performance through introduction of a dynamic inertia weight adjustment mechanism, an adaptive disturbance perception adjustment strategy and a role-layered collaborative optimization update mode, thereby having the advantages of strong adaptability, high control accuracy, good disturbance robustness and high optimization efficiency.

[0011] The particle swarm optimization-based offshore rocket recovery platform attitude stabilization control method according to an embodiment of the present application comprises the following steps:

[0012] S1, collecting multi-source attitude state data of an offshore rocket recovery platform, and preprocessing the multi-source attitude state data to generate an attitude state time series data set;

[0013] S2, constructing a fuzzy PID controller for controlling three control channels of pitch, roll and yaw attitude, each channel containing proportional gain, integral gain and derivative gain, a total of nine control parameters;

[0014] S3, using an improved particle swarm optimization algorithm to preliminarily globally optimize the nine control parameters of the fuzzy PID controller, and outputting an initial candidate solution set of control parameters;

[0015] S4, based on the initial candidate solution set of control parameters, introducing a zebra optimization algorithm to finely optimize the control parameters, and performing local improvement and compensation optimization for jumping out of local extreme value regions;

[0016] S5, constructing a collaborative optimization architecture integrating the improved particle swarm optimization algorithm and the zebra optimization algorithm, forming a master-slave solution space guidance mechanism, and outputting an optimal control parameter group after fusion optimization;

[0017] S6, inputting the optimal control parameter group into the fuzzy PID controller, combining the attitude state time series data set, driving the three-axis actuator in the servo system to real-time adjust the pitch angle, roll angle and yaw angle, and obtaining an attitude response result;

[0018] S7, monitoring attitude change data in the control process, and calculating a current disturbance amplitude index. When the disturbance amplitude exceeds a set threshold, a jumping behavior strategy in the zebra optimization algorithm is dynamically triggered, and the disturbance amplitude is fed back to the improved particle swarm optimization algorithm for adaptive adjustment of particle evolution behavior;

[0019] S8, comparing the attitude response result with the preset performance evaluation index, if the performance does not reach the set standard, re-executing steps S3 to S7 to form a closed-loop optimization process of adaptive updating of the control parameter.

[0020] Optionally, the multi-source attitude state data specifically includes pitch angle, roll angle, yaw angle, and corresponding angular velocity and acceleration, for providing fuzzy PID controller input state and disturbance identification basis.

[0021] Optionally, the preprocessing of the multi-source attitude state data specifically includes data denoising, time synchronization, and normalization processing, for improving stability of the multi-source attitude state data and accuracy of the fuzzy PID controller response.

[0022] Optionally, the S2 specifically includes:

[0023] S21, constructing a fuzzy PID controller model for controlling pitch angle, roll angle, and yaw angle of three-axis attitude, the fuzzy PID controller including three independent control channels corresponding to pitch channel, roll channel, and yaw channel respectively;

[0024] S22, each control channel containing proportional gain K pi , integral gain K ii , and differential gain K di , where i∈{1,2,3} respectively corresponding to pitch axis, roll axis, and yaw axis;

[0025] S23, calculating control output u i (t) of each channel, the control output of each channel being obtained by weighted calculation of the attitude error signal through the proportional gain, integral gain, and differential gain of the corresponding channel, for driving the attitude control actuator in the corresponding axis direction to realize real-time adjustment of the pitch angle, roll angle, and yaw angle;

[0026] S24, forming a control parameter vector K with nine control gain parameters of the three channels;

[0027] S25, taking the control parameter vector K as an optimization target, for determining the optimal controller parameter group by fusion of the improved particle swarm optimization algorithm and the zebra optimization algorithm.

[0028] Optionally, the S3 specifically includes:

[0029] S31, setting the particle swarm size as N, each particle representing a vector K composed of nine control parameters of the fuzzy PID controller, K = {K p1 ,K i1 ,K d1 ,K p2 ,K i2 ,K d2 ,Kp3 ,K i3 ,K d3 The particle's current position vector is X. j (t), velocity vector is V j (t), where j∈{1,2,...,N}, and N is the size of the particle swarm;

[0030] S32. Introduce a path backtracking hierarchical mechanism to record the position trajectory {X} of each particle within the most recent k iteration cycles. j (tk),...,X j X(t)}, and constructs a path similarity function based on the platform attitude response history sequence {y(tk),...,y(t)}, where X j (tk) represents the position vector of the control parameters of the j-th particle in the tk-th iteration, X j (t) represents the position vector of the control parameters of the j-th particle in the current t-th iteration, y(tk) represents the attitude response result after the controller output is completed in the tk-th control cycle, and y(t) represents the attitude response result after the controller output is completed in the t-th control cycle. Based on the similarity score, the particles are divided into three categories: path following subgroup, path jumping subgroup, and path prediction subgroup, and different search behavior rules are applied to each category.

[0031] S33. Set the platform multidimensional disturbance intensity vector δ(t)=[δ θ (t),δ φ (t),δ ψ [(t)], calculate the perturbation mode length ‖δ(t)‖, and dynamically adjust the inertia weight ω(t) through a piecewise linear function to adapt to the particle search capability adjustment under different perturbation levels:

[0032]

[0033] Where, δ θ (t) represents the disturbance intensity in the pitch direction during the t-th control cycle, δ φ (t) represents the disturbance intensity in the roll direction during the t-th control cycle, δ ψ (t) represents the disturbance intensity in the yaw angle direction during the t-th control cycle, where ω1, ω2 and ω3 are three preset inertial weight constants, and ω1>ω2>ω3, and ∈1 and ∈2 are the disturbance intensity segmentation threshold parameters;

[0034] S34. Construct a fuzzy logic adjustment mechanism and set the attitude error change rate. With the local convergence amplitude ρ of the particle jFor fuzzy input, set self-learning factor c1(t) and group learning factor c2(t) as fuzzy output, and adjust the learning factor through fuzzy rule base, so as to realize adaptive regulation and control of particle search ability based on platform control behavior feedback;

[0035] S35, update the speed and position through the difference vector between the current state of the fusion particle and the individual optimal position and the group optimal position, so as to guide the particle to dynamically adjust the direction in the search space and evolve to the optimal control parameter:

[0036] V j (t+1)=ω(t)·V j (t)+c1(t)·r1·(P j (t)-X j (t))+c2(t)·r2·(G(t)-X j (t));

[0037] X j (t+1)=X j (t)+V j (t+1);

[0038] Wherein, V j (t+1) represents the updated speed vector of the jth particle at the t+1th iteration, V j (t) represents the updated speed vector of the jth particle at the tth iteration, r1, r2 are random numbers, P j (t) is the individual optimal position of the particle, G(t) is the current global optimal position, X j (t+1) is the updated position vector of the jth particle at the t+1th iteration;

[0039] S36, evaluate the performance of the control parameter combination obtained in each iteration by the multi-objective fitness function and update P j (t) and G(t), and output the control parameter initial candidate solution set when the maximum iteration number is reached or the convergence condition is met.

[0040] Optionally, the S4 specifically comprises:

[0041] S41, taking the output control parameter initial candidate solution set {K1, K2,..., K M} as the initial population of zebra optimization algorithm, wherein each control parameter individual

[0042] m∈{1,2,...,M}, M represents the total number of control parameter individuals, which is used to constitute a complete fuzzy PID controller control parameter group;

[0043] S42, construct a leader score function L based on the attitude response data m , for measuring the comprehensive ability of control parameter individuals in stability and disturbance adaptability, and performing particle role division:

[0044]

[0045] Where λ is a weighting factor, STD(y m (t)) is the standard deviation of the platform attitude response sequence y m (t), T s,m is the attitude convergence time corresponding to the mth control parameter individual, Δy m is the change amplitude of the platform attitude deviation caused by the mth control parameter individual, Δt m is the disturbance recovery time required for the attitude deviation to recover from the maximum value to the steady state range;

[0046] S43, sort all individuals in descending order of the leader score function L m , and select the top P% individuals as the leader individual set The rest are as the follower individual set And respectively apply a differentiated search strategy to perform position update operation on each follower individual:

[0047] K m (t+1)=K m (t)+α·rand()·(K best (t)-K m (t))+β·randn();

[0048] Where K m (t+1) is the updated position vector of the mth control parameter individual after the t+1th iteration, K m (t) is the updated position vector of the mth control parameter individual after the tth iteration, K best (t) is the current global optimal control parameter individual, α is the zebra global tracking coefficient, β is the local disturbance coefficient, rand() represents a uniform random number in the interval [0, 1], and randn() represents a normal distribution disturbance term;

[0049] S44, apply an escape update mechanism to the leader individual to avoid falling into a local extremum, and perform the following escape update:

[0050] K lead (t+1)=K lead (t)+γ·sign(K lead (t)-K mean (t))·D;

[0051] Where γ is the escape intensity coefficient, Kmean (t) is the current iteration population average control parameter vector, D is the preset maximum disturbance amplitude, sign() is the sign function, K lead (t+1) is the leader individual control parameter vector after escape update, K lead (t) is the leader individual control parameter vector at the tth iteration;

[0052] S45, repeatedly performing S43 to S44, the number of iterations reaches the set upper limit T z or the optimal fitness convergence threshold, output the final optimal control parameter group K * , used for fuzzy PID controller parameter input.

[0053] Optionally, the S5 specifically includes:

[0054] S51, establish a global search module in solution space dominated by improved particle swarm optimization algorithm, for generating multiple control parameter initial candidate solutions and guiding the overall evolution direction of solution space;

[0055] S52, construct a local fine search module controlled by zebra optimization algorithm, perturb and guide the candidate solutions generated by the master module, and enhance the local optimization ability and the ability to jump out of local optimal trap;

[0056] S53, establish an information interaction mechanism between the master and slave modules, periodically feed back the local optimal solution in the zebra optimization algorithm to the particle swarm optimization algorithm, for adjusting the global guide solution of the particle swarm and updating the search direction;

[0057] S54, based on the performance of the control parameter individual in the improved particle swarm optimization algorithm and the zebra optimization algorithm, construct a multi-strategy joint scheduling module, dynamically determine the search path suitable for global expansion or local convergence of each control parameter solution, and respectively allocate to the master and slave modules for iterative update;

[0058] S55, construct a fusion optimization parameter evaluation mechanism, comprehensively evaluate the fitness of the control parameter group output by each round of master-slave collaborative search, and extract the optimal control parameter group as the fusion optimization result;

[0059] S56, through the dynamic change trend in the fusion evaluation result, feedback adjust the information exchange frequency between the master-slave modules and the control parameter selection strategy, realize performance-driven collaborative optimization scheduling and structural adaptive adjustment.

[0060] Optionally, the S6 specifically includes:

[0061] S61, input the optimal control parameter group after fusion optimization to the fuzzy PID controller, for configuring the proportional, integral and derivative gain parameters of the three channels of the controller;

[0062] S62, read the real-time attitude input of the current platform from the attitude state time series data set, including the state values of the pitch angle, roll angle and yaw angle, as the input reference of the fuzzy PID controller;

[0063] S63, the control output of the three attitude channels is calculated by the fuzzy PID controller, which corresponds to the error signal adjustment results of the pitch, roll and yaw directions respectively;

[0064] S64, the control signal output by the controller is transmitted to the servo system in real time, which drives the three-axis actuator to act on the attitude adjustment of the pitch, roll and yaw directions respectively, and the servo system is specifically composed of three-axis electric actuator, servo driver, control signal receiving module and feedback sensor module, which is used to receive control instructions, drive platform attitude adjustment and real-time feedback execution state;

[0065] S65, the servo system performs corresponding physical actions on the platform structure according to the received control instructions, completes the real-time correction and adjustment operation of the attitude;

[0066] S66, after the platform attitude adjustment is completed, the latest response state data of each axis is collected, the attitude response result is formed, and is fed back for performance evaluation and dynamic optimization link.

[0067] The beneficial effects of the present application are:

[0068] The present application successfully realizes the global-local linkage optimization of the nine control parameters of the fuzzy PID controller by constructing the master-slave collaborative optimization architecture of the improved particle swarm optimization algorithm and the zebra optimization algorithm, solves the key problems of inaccurate parameter adjustment, large control performance fluctuation, algorithm convergence precocity and weak anti-disturbance ability in the traditional optimization method. Compared with the single particle swarm algorithm optimization method in the prior art, the present application introduces the population stratification mechanism driven by the leader score, the adaptive disturbance perception adjustment strategy and the mutation probability feedback mechanism in the optimization structure, so that the particle swarm search behavior can dynamically adjust the search direction and range according to the platform attitude disturbance state, thereby improving the stability and global search ability of the optimization.

[0069] The information interaction and cooperative scheduling between the master particle swarm algorithm and the slave zebra optimization algorithm solve the algorithm conflict and structural redundancy problems in the fusion optimization, and significantly enhance the adaptability and practicality of parameter optimization. Specifically, the particle swarm provides a global search dominant path, and the zebra optimization performs a jumping disturbance and strategy guidance in a local area, so that the fusion optimization result has both convergence accuracy and the ability to jump out of the local optimum. Meanwhile, the fuzzy PID controller is input combined with the attitude state time series data and the optimal control parameter group to drive the three-axis actuator of the servo system to realize the rapid and stable adjustment of the pitch angle, roll angle and yaw angle in complex sea conditions, and good attitude response results are obtained.

[0070] Finally, the method proposed in the application can realize closed-loop cooperative evolution between controller parameter optimization and platform response performance, so that the attitude adjustment process reaches optimal configuration in response time, stability, energy consumption control and other multi-dimensional performance indicators, has strong dynamic environment adaptability, high attitude adjustment precision, strong control robustness and high system self-adaptive optimization efficiency, significantly improves the attitude stabilization control ability of the offshore rocket recovery platform, and has good engineering practical value and popularization prospect. BRIEF DESCRIPTION OF DRAWINGS

[0071] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, illustrate the application together with the embodiments thereof, and explain the application, and do not constitute a limitation on the application. In the drawings:

[0072] Figure 1 The flowchart of the offshore rocket recovery platform attitude stabilization control method based on particle swarm optimization proposed in the application is shown in the figure.

[0073] Figure 2 The zebra optimization algorithm disturbance update and role grouping mechanism diagram of the offshore rocket recovery platform attitude stabilization control method based on particle swarm optimization proposed in the application is shown in the figure. DETAILED DESCRIPTION

[0074] The application will now be described in further detail with reference to the drawings. These drawings are simplified schematic diagrams, and only illustrate the basic structure of the application in a schematic manner, and therefore only show the components related to the application.

[0075] REFERENCE Figure 1 and Figure 2 The offshore rocket recovery platform attitude stabilization control method based on particle swarm optimization includes the following steps:

[0076] S1, collect multi-source attitude state data of the offshore rocket recovery platform, and pre-process the multi-source attitude state data to generate an attitude state time series data set;

[0077] S2, a fuzzy PID controller is constructed for controlling three control channels of pitch, roll and yaw three-axis attitude, each channel containing proportional gain, integral gain and derivative gain, a total of nine control parameters;

[0078] S3, the improved particle swarm optimization algorithm is used to preliminarily globally optimize the nine control parameters of the fuzzy PID controller, and an initial candidate solution set of the control parameters is output;

[0079] S4, based on the initial candidate solution set of the control parameters, the zebra optimization algorithm is introduced to finely optimize the control parameters, and the control parameters are locally improved and compensated for jumping out of the local extreme value region;

[0080] S5, a collaborative optimization architecture of the improved particle swarm optimization algorithm and the zebra optimization algorithm is constructed, a master-slave solution space guiding mechanism is formed, and an optimal control parameter group after fusion optimization is output;

[0081] S6, the optimal control parameter group is input into the fuzzy PID controller, combined with the attitude state time series data set, the three-axis actuator in the servo system is driven to adjust the pitch angle, roll angle and yaw angle in real time, and an attitude response result is obtained;

[0082] S7, the attitude change data in the control process is monitored, and the current disturbance amplitude index is calculated, when the disturbance amplitude exceeds the set threshold, the jumping behavior strategy in the zebra optimization algorithm is dynamically triggered, and the disturbance amplitude is fed back to the improved particle swarm optimization algorithm for adaptive adjustment of the particle evolution behavior;

[0083] S8, based on the comparison between the attitude response result and the preset performance evaluation index, if the performance does not reach the set standard, steps S3 to S7 are re-executed, and a closed-loop optimization process of adaptive updating of the control parameters is formed.

[0084] By constructing the collaborative optimization architecture of the particle swarm optimization algorithm and the zebra optimization algorithm, the adaptive and efficient setting of the nine control parameters of the fuzzy PID controller is realized, and the attitude stability control ability of the offshore rocket recovery platform in complex sea conditions is significantly improved. Compared with the traditional manual parameter adjustment or single optimization algorithm, the master-slave solution space guiding mechanism is introduced, the global search is performed by the particle swarm algorithm, the local disturbance fine optimization is responsible by the zebra algorithm, and the jumping strategy is driven by the disturbance amplitude monitoring, so that the dynamic adaptation of the control performance and the disturbance response ability is realized. At the same time, the closed-loop optimization feedback mechanism is introduced, the parameter optimization process is dynamically adjusted based on the comparison between the attitude response and the performance index, and the system always maintains high precision, fast response and low energy consumption in the variable environment. The present application has the advantages of high control precision, fast convergence speed, strong anti-disturbance ability and high intelligentization degree of optimization process, and significantly improves the control stability and engineering applicability of the rocket recovery platform in the dynamic disturbance scene.

[0085] In the embodiment, the multi-source attitude state data specifically includes pitch angle, roll angle, yaw angle, and corresponding angular velocity and acceleration, which are used to provide fuzzy PID controller input state and disturbance identification basis.

[0086] In the embodiment, the preprocessing of the multi-source attitude state data specifically includes data denoising, time synchronization, and normalization processing, which are used to improve the stability of the multi-source attitude state data and the accuracy of the fuzzy PID controller response.

[0087] In the embodiment, the S2 specifically includes:

[0088] S21, a fuzzy PID controller model for controlling the pitch angle, roll angle, and yaw angle of the three-axis attitude is constructed, and the fuzzy PID controller includes three independent control channels corresponding to the pitch channel, roll channel, and yaw channel, respectively;

[0089] S22, each control channel includes proportional gain K pi , integral gain K ii , and differential gain K di , where i∈{1,2,3} respectively correspond to the pitch axis, roll axis, and yaw axis;

[0090] S23, the control output u i (t) of each channel is calculated, which is obtained by weighting the attitude error signal through the proportional gain, integral gain, and differential gain of the corresponding channel, and is used to drive the attitude control actuator in the corresponding axis direction to realize real-time adjustment of the pitch angle, roll angle, and yaw angle;

[0091] S24, nine control gain parameters of the three channels form a control parameter vector K;

[0092] S25, the control parameter vector K is used as an optimization target to determine the optimal controller parameter set by the fusion of the improved particle swarm optimization algorithm and the zebra optimization algorithm.

[0093] The application realizes fine channel control of three-axis attitude of pitch angle, roll angle and yaw angle by constructing a three-axis independent fuzzy PID controller model, and fully embodies the design idea of multi-channel cooperative control and parameter structure decoupling. Each control channel respectively introduces proportional gain, integral gain and differential gain, so that the controller can consider rapidity, stability and steady-state accuracy in the response of attitude error, effectively improving the response sensitivity and robustness of platform attitude adjustment. By uniformly constructing nine control gain parameters as a control parameter vector, the system has high parameterization and optimization, which is convenient for seamless docking with swarm intelligence algorithm. On this basis, the control parameter vector is taken as the target input of the fusion optimization algorithm, which can effectively guide the particle swarm and zebra optimization algorithm to jointly search for the optimal control parameter configuration in the global and local range, improve the consistency and adjustability of the attitude control performance. The structure not only strengthens the independent adjustment ability of each axis control channel, but also improves the coordination of multi-axis linkage control, ensures that the rocket recovery platform realizes high-precision and high-stability attitude control in the dynamic disturbance environment, and has significant engineering practical value and system optimization advantage.

[0094] In the embodiment, the S3 specifically includes:

[0095] S31, set the particle swarm size to N, and each particle represents a vector K = {K p1 ,K i1 ,K d1 ,K p2 ,K i2 ,K d2 ,K p3 ,K i3 ,K d3} composed of nine control parameters of the fuzzy PID controller, the current position vector of the particle is X j (t), and the velocity vector is V j (t), wherein j is in {1, 2,..., N}, N is the size of the particle swarm;

[0096] S32, a path backtracking hierarchical mechanism is introduced, the position trajectory {X j (t-k),..., X j (t)} of each particle in the last k iteration periods is recorded, and a path similarity function is constructed based on the historical sequence {y(t-k),..., y(t)} of platform attitude response, wherein X j (t-k) represents the control parameter position vector of the jth particle at the t-kth iteration, and X j(t) represents the control parameter position vector of the jth particle at the current tth iteration, y(t-k) represents the attitude response result after the controller output at the t-kth control period is completed, y(t) represents the attitude response result after the controller output at the tth control period is completed, the particles are divided into three categories of path following subgroups, path jumping subgroups and path prediction subgroups according to the similarity scores, and different search behavior rules are applied respectively;

[0097] S33, set the platform multi-dimensional disturbance intensity vector δ(t) = [δ θ (t), δ φ (t), δ ψ (t)], calculate the disturbance module length ‖δ(t)‖, and dynamically adjust the inertia weight ω(t) through the broken line segmentation function, for adjusting the particle search ability under different disturbance levels:

[0098]

[0099] Wherein, δ θ (t) represents the disturbance intensity in the pitch angle direction in the tth control period, δ φ (t) represents the disturbance intensity in the roll angle direction in the tth control period, δ ψ (t) represents the disturbance intensity in the yaw angle direction in the tth control period, ω1, ω2 and ω3 are three preset inertia weight constants, and ω1> ω2> ω3, ∈1 and ∈2 are disturbance intensity segmentation threshold parameters;

[0100] S34, construct a fuzzy logic regulation mechanism, set the attitude error change rate and the local convergence amplitude of the particle ρ j as fuzzy inputs, set the self-learning factor c1(t) and the group learning factor c2(t) as fuzzy outputs, adjust the learning factors through the fuzzy rule base, and realize the adaptive regulation of the particle search ability based on the platform control behavior feedback;

[0101] S35, update the velocity and position by fusing the difference vectors between the current state of the particle and the individual optimal position and the group optimal position, for guiding the particle to dynamically adjust the direction in the search space and evolve to the optimal control parameter:

[0102] V j (t+1) = ω(t)·V j (t) + c1(t)·r1·(P j (t) - X j (t)) + c2(t)·r2·(G(t) - X j (t));

[0103] X j (t+1) = Xj (t)+V j (t+1);

[0104] Wherein, V j (t+1) represents the updated speed vector of the jth particle at the t+1th iteration, V j (t) represents the updated speed vector of the jth particle at the tth iteration, r1, r2 are random numbers, P j (t) is the individual optimal position of the particle, G(t) is the current global optimal position, X j (t+1) is the updated position vector of the jth particle at the t+1th iteration;

[0105] S36, the performance of each iteration obtained control parameter combination is evaluated by a multi-objective fitness function, and P j (t) and G(t), when the maximum number of iterations is reached or the convergence condition is met, the initial candidate solution set of the control parameters is output.

[0106] The present application improves the structure of the particle swarm optimization algorithm, introduces a path backtracking hierarchical mechanism, a disturbance perception dynamic weight adjustment strategy and a fuzzy learning factor adjustment method, so that the particle swarm search behavior is more environmentally adaptive and has a control feedback response ability. Through the path backtracking mechanism, the evolution trajectory of the particle in the historical iteration can be effectively evaluated, and the path similarity is calculated combined with the platform attitude response data, the intelligent grouping and differentiated search regulation of the particle are realized, so that the ability of the algorithm to jump out of the local optimum is improved. At the same time, aiming at the multi-dimensional disturbance characteristics of the offshore platform, a disturbance intensity vector is constructed and the inertia weight is dynamically adjusted based on a piecewise function, so that the particle search ability is closely coupled with the current disturbance state, and the robustness to different working conditions is improved. Further, a fuzzy logic adjustment mechanism is introduced, taking the attitude error change rate and the convergence amplitude as inputs, and dynamically adjusting the particle learning factor, so that the search behavior has higher autonomy and control feedback sensitivity. Finally, through an adaptive updating strategy, the search path is guided by individual and group experience, and a more reasonable search updating mechanism is constructed. In summary, the present application significantly enhances the search efficiency and convergence accuracy of the particle swarm in the optimization of nonlinear and highly disturbed platform control, and provides more intelligent and stable global optimization support for fuzzy PID parameter tuning.

[0107] In the embodiment, the S4 specifically comprises:

[0108] S41, the initial candidate solution set of the control parameters output {K1, K2,..., K M} is taken as the initial population of the zebra optimization algorithm, wherein each control parameter individual

[0109] m e {1, 2, …, M}, M represents the total number of control parameter individuals, which is used to constitute a complete fuzzy PID controller control parameter group;

[0110] S42, based on the attitude response data, a leader score function L is constructed m , which is used to measure the comprehensive ability of the control parameter individuals in stability and disturbance adaptability, and the particle role is divided:

[0111]

[0112] Where λ is a weighting factor, STD(y m (t)) is the standard deviation of the platform attitude response sequence y m (t), T s,m is the attitude convergence time corresponding to the mth control parameter individual, Δy m is the change amplitude of the platform attitude deviation caused by the mth control parameter individual, Δt m is the disturbance recovery time required for the attitude deviation to recover from the maximum value to the steady state range;

[0113] S43, sort all individuals in descending order according to the leader score function L m , select the top P% individuals as the leader individual set The rest is the follower individual set And respectively apply differential search strategy, execute position update operation to each follower individual:

[0114] K m (t+1) = K m (t) + α·rand()·(K best (t) - K m (t)) + β·randn();

[0115] Where K m (t+1) is the updated position vector of the mth control parameter individual after the t+1th iteration, K m (t) is the updated position vector of the mth control parameter individual after the tth iteration, K best (t) is the current global optimal control parameter individual, α is the zebra global tracking coefficient, β is the local disturbance coefficient, rand() represents a uniform random number in the interval [0, 1], and randn() represents a normal distribution disturbance term;

[0116] S44, apply escape update mechanism to the leader individual to avoid falling into local extremum, execute the following escape update:

[0117] K lead (t+1) = K lead (t) + γ·sign(Klead (t) - K mean (t) · D

[0118] wherein, gamma is an escape intensity coefficient, K mean (t) is a current iteration population average control parameter vector, D is a preset maximum disturbance amplitude, sign() is a sign function, K lead (t+1) is a leader individual control parameter vector after escape update, K lead (t) is a leader individual control parameter vector at the tthiteration;

[0119] S45, repeatedly performing S43 to S44, the iteration number reaches the set upper limit T z or the optimal fitness convergence threshold, output the final optimized control parameter group K * , for fuzzy PID controller parameter input.

[0120] The present application is aimed at the fine optimization stage of control parameters, and a zebra optimization algorithm architecture based on role division and behavior differentiation is constructed. The posture control performance of the control parameter individual is comprehensively evaluated by the leader scoring function, effectively solving the problems of fuzzy role division, insufficient global guidance and weak local jumping ability in traditional optimization algorithms. The scoring function simultaneously considers the stability index and disturbance adaptability of the platform posture response, making the individual layering more physically meaningful. After role division based on the scoring results, the leader individual executes the escape update strategy, effectively breaking through the local extreme region and enhancing the global jumping ability. The follower individual executes dynamic update combining global guidance and local disturbance, realizing robust convergence. In addition, the constructed zebra global tracking coefficient and disturbance factor are coupled to search the path, and the normal disturbance is introduced to enhance the local search breadth and improve the solution space coverage. Through multiple iterations and fitness feedback adjustment, the final optimized parameter group has higher control stability and robustness in the fuzzy PID controller. In summary, the present application significantly improves the local tuning precision and posture recovery speed of the control parameters, enhances the global adaptability and dynamic response ability of the system, and provides an efficient and intelligent optimization support scheme for the fuzzy control system.

[0121] In the embodiment, the S5 specifically includes:

[0122] S51, a solution space global search module controlled by the improved particle swarm optimization algorithm is established, for generating a plurality of control parameter initial candidate solutions and guiding the overall evolution direction of the solution space;

[0123] S52, a local fine search module controlled by the zebra optimization algorithm is constructed, for disturbing and guiding the candidate solutions generated by the main control module and compensating for the jumping, enhancing the local optimization ability and the ability to jump out of the local optimal trap;

[0124] S53, an information interaction mechanism is established between the master module and the slave module, and a local optimal solution in the zebra optimization algorithm is fed back to the particle swarm optimization algorithm periodically, so that the global guiding solution of the particle swarm is adjusted and the search direction is updated;

[0125] S54, based on the performance of the control parameter individual in the improved particle swarm optimization algorithm and the zebra optimization algorithm, a multi-strategy joint scheduling module is constructed, and it is dynamically judged that each control parameter solution is suitable for the search path of global expansion or local convergence, and is respectively dispatched to the master module and the slave module for iterative updating;

[0126] S55, a fusion optimization parameter evaluation mechanism is constructed, the fitness of the control parameter group output by each round of master-slave collaborative search is comprehensively evaluated, and the optimal control parameter group is extracted as the fusion optimization result;

[0127] S56, through the dynamic change trend in the fusion evaluation result, the information exchange frequency and the control parameter selection strategy between the master module and the slave module are fed back and adjusted, and the performance-driven collaborative optimization scheduling and structural adaptive adjustment are realized.

[0128] The improved particle swarm optimization algorithm and the zebra optimization algorithm are constructed in the master-slave collaborative optimization architecture, the efficient fusion of global search ability and local fine optimization ability is realized, and the problems of single search direction, easy to fall into local optimum and slow global convergence of traditional single optimization algorithm are broken through. The particle swarm as the master module undertakes the initial candidate solution generation and solution space evolution task of the control parameter, and guarantees the breadth and coverage of global search. The zebra optimization algorithm as the slave module is responsible for the jump disturbance and local fine compensation of the master candidate solution, and the precision and robustness of the optimization result are improved. A periodic interaction mechanism is established between the master module and the slave module, so that the high-quality local solution in the zebra optimization can feed back to the particle swarm evolution path, and the group guiding strategy is effectively adjusted. The dynamic scheduling mechanism constructed in combination with the parameter performance enables different solutions to select appropriate search paths according to the fitness performance, improves the pertinence and collaboration of the optimization process. At the same time, through the fusion of fitness evaluation mechanism and dynamic feedback strategy, the system can realize the adaptive adjustment of the information interaction frequency and the scheduling strategy between the master-slave structure, and enhance the flexibility and intelligent level of the optimization architecture. In summary, the present application greatly improves the efficiency, precision and stability of the fuzzy PID controller parameter optimization, and is suitable for precise attitude control requirements of high disturbance and nonlinear complex systems.

[0129] In the embodiment, the S6 specifically comprises:

[0130] S61, input the optimal control parameter group after fusion optimization to the fuzzy PID controller, for configuring the proportional, integral and differential gain parameters of the three channels of the controller;

[0131] S62, read the real-time attitude input of the current platform from the attitude state time series data set, including the state values of the pitch angle, roll angle and yaw angle, as the input reference of the fuzzy PID controller;

[0132] S63, the control output of the three attitude channels is calculated by the fuzzy PID controller, which corresponds to the error signal adjustment results of the pitch, roll and yaw directions respectively;

[0133] S64, the control signal output by the controller is transmitted to the servo system in real time to drive the three-axis actuator to act on the pitch, roll and yaw directions for attitude adjustment, and the servo system is specifically composed of three-axis electric actuators, servo drivers, control signal receiving modules and feedback sensor modules, for receiving control instructions, driving platform attitude adjustment and real-time feedback of execution state;

[0134] S65, the servo system performs corresponding physical actions on the platform structure according to the received control instructions to complete the real-time correction and adjustment of the attitude;

[0135] S66, after the platform attitude adjustment is completed, the latest response state data of each axis is collected to form the attitude response result, which is fed back for performance evaluation and dynamic optimization.

[0136] The optimal control parameter group after fusion optimization is input into the fuzzy PID controller, so as to realize the precise configuration of the proportional, integral and differential gain parameters of the three-channel (pitch, roll and yaw) attitude control, greatly improve the response speed and adjustment accuracy of the control system to complex disturbances. By connecting the real-time input state in the attitude state time series data set, the fuzzy PID controller can perform error calculation and control output generation based on the current platform dynamic attitude information, so as to ensure that the control instruction is close to the actual running state. The controller output signal drives the servo system in real time, and the three-axis electric actuators jointly complete the physical adjustment operation of the platform attitude, and cooperate with the feedback sensor module to realize the closed loop feedback of the attitude execution state, so as to ensure accurate adjustment and efficient execution. The structure makes the control system have the ability of real-time regulation, feedback correction and execution synchronization, which significantly improves the stability and tracking performance of the platform in the wind and wave disturbance environment. At the same time, by collecting the platform response state to form the attitude response result, the performance evaluation module and the subsequent parameter optimization can be provided with data basis, and a performance-driven closed-loop regulation mechanism is constructed. In summary, this link ensures the execution of the control strategy while strengthening the data feedback chain and self-adaptive updating ability of the control system, providing full-link technical support and engineering support for platform attitude stable control.

[0137] Example 1:

[0138] In order to verify the feasibility of the application in practice, the application is applied to a certain floating recovery test platform, which is provided with a three-axis electric actuator, an inertial measurement unit (IMU), a main control calculation module and a fuzzy PID controller unit, and has a recovery section attitude adjustment capability. The test target is to compare the response accuracy, adjustment speed and energy consumption of the platform attitude control system before and after optimization under typical sea state disturbance conditions, and to evaluate the application performance of the application technology in actual complex environment.

[0139] In the test environment, the platform withstands wind and wave disturbance frequency of 0.12-0.18 Hz, and the average wave height is about 0.8 meters. The test first uses the traditional method of manually setting fuzzy PID parameters to control the attitude, and records the maximum control deviation, control response convergence time and average control energy consumption of the platform in the pitch angle, roll angle and yaw angle three-axis direction. The results show that the system response is slow, easy to overshoot or oscillate, the platform stability is poor, the control convergence time is close to 9 seconds on average, and the control power consumption is high.

[0140] Subsequently, the control method described in the application is enabled for comparison test. First, the master-slave collaborative optimization architecture is constructed by improving the particle swarm optimization algorithm combined with the zebra optimization algorithm, the control parameter candidate set is constructed based on historical control data and real-time attitude feedback, and the disturbance perception feedback, role layering mechanism and adaptive mutation strategy are introduced in the search process, and the nine fuzzy PID control parameters are dynamically optimized and configured. After optimization, the controller is imported for real-time control, and the attitude response data is continuously collected and analyzed and compared.

[0141] According to the comparison results, the three-axis maximum attitude deviation of the optimized platform is greatly reduced, among which the pitch angle is reduced from 5.72° to 2.14°, the roll angle is reduced from 4.81° to 1.87°, and the yaw angle is reduced from 3.49° to 1.23°. The control response speed is also significantly improved, and the convergence time is shortened from 8.96 seconds to 3.76 seconds, with an efficiency improvement of more than 58%. At the same time, due to the optimization of the control parameters, the control accuracy and system stability are improved, and the control energy consumption is reduced from 143.55 kJ to 96.87 kJ, with obvious energy saving effect.

[0142] Table 1 Comparison of key performance indicators of offshore platform attitude control

[0143] Performance index Before optimization After optimization Maximum pitch angle deviation (°) 5.72 2.14 Maximum roll angle deviation (°) 4.81 1.87 Maximum yaw angle deviation (°) 3.49 1.23 Attitude control convergence time (s) 8.96 3.76 Average control energy consumption (kJ) 143.55 96.87

[0144] According to the data in Table 1, it can be clearly seen that the attitude stability control method proposed in the application has achieved significant optimization in multiple key performance indicators compared with the traditional control method. The table reflects the essential difference between the system before and after optimization in actual control performance, starting from three main attitude control indicators of pitch angle, roll angle and yaw angle, combining convergence time and average control energy consumption, a total of five performance indicators are compared.

[0145] Firstly, in terms of maximum pitch angle deviation, the deviation under traditional control mode is as high as 5.72°, which is prone to cause the platform to tilt forward or backward unstably in strong fluctuation sea conditions, and there is a risk of rocket landing. After introducing the method of the application, the index is significantly reduced to 2.14°, and the control accuracy is improved by more than 60%, which shows that the fuzzy PID controller parameters are more adaptive to the disturbance dynamics under the action of swarm intelligence optimization, effectively suppressing overshoot and oscillation.

[0146] The same trend is also reflected in the roll angle control index. The maximum deviation before optimization is 4.81°, and it is reduced to 1.87° after optimization, which is crucial in dealing with disturbances such as side wave impact. The reduction of yaw angle error is particularly obvious, from 3.49° to 1.23°, which shows that the fusion optimization algorithm has stronger synchronous optimization ability in three-axis linkage balance adjustment, and the overall coordination of the system posture is improved.

[0147] In terms of dynamic response capability, the posture convergence time before optimization is 8.96 seconds, and it is shortened to 3.76 seconds after optimization, with a response speed improvement of more than 58%. This result shows that the control system responds more quickly to disturbance changes, and the control loop calculation logic and execution link are significantly strengthened, which is suitable for high-speed change scenarios such as rocket descent stage.

[0148] Finally, the average control energy consumption is reduced from 143.55kJ to 96.87kJ, which shows that the optimized control strategy is not only more accurate, but also more energy-efficient. The parameter setting is more reasonable, the controller output is stable, and unnecessary high-frequency regulation actions are avoided, reducing the energy consumption burden of the actuator, and having good engineering economy.

[0149] In summary, the data in the table directly proves the significant advantages of the application in offshore platform posture stability control. Whether it is control accuracy, response speed or energy efficiency, each core index has been comprehensively improved, verifying the efficiency and practicality of the improved particle swarm optimization algorithm and zebra optimization algorithm fusion mechanism in complex environments, and reflecting the engineering promotion value and performance leading advantage of the application.

[0150] The above is only the preferred specific embodiment of the application, but the protection scope of the application is not limited thereto, and any skilled person in the art can make equivalent substitutions or changes to the technical solutions and inventive concepts of the application within the technical scope disclosed by the application, which should be covered within the protection scope of the application.

Claims

1. A method for attitude stabilization control of a marine rocket recovery platform based on particle swarm optimization, characterized in that, Includes the following steps: S1. Collect multi-source attitude state data of the marine rocket recovery platform, and preprocess the multi-source attitude state data to generate an attitude state time series dataset. S2. Construct a fuzzy PID controller to control three control channels for pitch, roll and yaw attitude. Each channel contains proportional gain, integral gain and derivative gain, for a total of nine control parameters. S3. An improved particle swarm optimization algorithm is used to perform preliminary global optimization on the nine control parameters of the fuzzy PID controller, and the initial candidate solution set of the control parameters is output. S4. Based on the initial candidate solution set of control parameters, the Zebra Optimization Algorithm is introduced to refine the control parameters, and the control parameters are locally improved and compensated to escape the local extreme value region. S5. Construct a collaborative optimization architecture that integrates the improved particle swarm optimization algorithm and the zebra optimization algorithm, form a master-slave solution space guidance mechanism, and output the optimal control parameter set after fusion optimization. S6. Input the optimal control parameter set into the fuzzy PID controller, and combine it with the attitude state time series dataset to drive the three-axis actuator in the servo system to adjust the pitch angle, roll angle and yaw angle in real time to obtain the attitude response result. S7. Monitor attitude change data during the control process and calculate the current disturbance amplitude index. When the disturbance amplitude exceeds the set threshold, dynamically trigger the jump behavior strategy in the Zebra optimization algorithm, and at the same time feed back the disturbance amplitude to the improved particle swarm optimization algorithm for adaptive adjustment of particle evolution behavior. S8. Based on the attitude response results, compare them with the preset performance evaluation indicators. If the performance does not meet the set standard, repeat steps S3 to S7 to form a closed-loop optimization process for adaptive updating of control parameters.

2. The attitude stabilization control method for a marine rocket recovery platform based on particle swarm optimization according to claim 1, characterized in that, The multi-source attitude state data specifically includes pitch angle, roll angle, yaw angle, and corresponding angular velocity and acceleration, which are used to provide the input state and disturbance identification basis for the fuzzy PID controller.

3. The attitude stabilization control method for a marine rocket recovery platform based on particle swarm optimization according to claim 1, characterized in that, The preprocessing of multi-source attitude state data specifically includes data denoising, time synchronization, and normalization, which are used to improve the stability of multi-source attitude state data and the accuracy of the fuzzy PID controller response.

4. The attitude stabilization control method for a marine rocket recovery platform based on particle swarm optimization according to claim 1, characterized in that, S2 specifically includes: S21. Construct a fuzzy PID controller model for controlling the three-axis attitude of pitch angle, roll angle and yaw angle. The fuzzy PID controller includes three independent control channels, corresponding to the pitch channel, roll channel and yaw channel respectively. S22, Each control channel includes a proportional gain K pi Integral gain K ii and differential gain K di , where i∈{1,2,3}, correspond to the pitch axis, roll axis and yaw axis respectively; S23. Calculate the control output u for each channel. i (t), the control output of each channel is obtained by weighting the attitude error signal with the proportional gain, integral gain and derivative gain of the corresponding channel, and is used to drive the attitude control actuator of the corresponding axis to realize real-time adjustment of pitch, roll and yaw angles; S24. The nine control gain parameters of the three channels are combined to form a control parameter vector K; S25. The control parameter vector K is used as the optimization objective and is used to determine the optimal set of controller parameters by the fusion of the improved particle swarm optimization algorithm and the zebra optimization algorithm.

5. The attitude stabilization control method for a marine rocket recovery platform based on particle swarm optimization according to claim 1, characterized in that, S3 specifically includes: S31. Set the particle swarm size to N, where each particle represents a vector K = {K} consisting of the nine control parameters of the fuzzy PID controller. p1 ,K i1 ,K d1 ,K p2 ,K i2 ,K d2 ,K p3 ,K i3 ,K d3 The particle's current position vector is X. j (t), velocity vector is V j (t), where j∈{1,2,...,N}, and N is the size of the particle swarm; S32. Introduce a path backtracking hierarchical mechanism to record the position trajectory {X} of each particle within the most recent k iteration cycles. j (tk),...,X j X(t)}, and constructs a path similarity function based on the platform attitude response history sequence {y(tk),...,y(t)}, where X j (tk) represents the position vector of the control parameters of the j-th particle in the tk-th iteration, X j (t) represents the position vector of the control parameters of the j-th particle in the current t-th iteration, y(tk) represents the attitude response result after the controller output is completed in the tk-th control cycle, and y(t) represents the attitude response result after the controller output is completed in the t-th control cycle. Based on the similarity score, the particles are divided into three categories: path following subgroup, path jumping subgroup, and path prediction subgroup, and different search behavior rules are applied to each category. S33. Set the platform multidimensional disturbance intensity vector δ(t)=[δ θ (t),δ φ (t),δ ψ [(t)], calculate the perturbation mode length ‖δ(t)‖, and dynamically adjust the inertia weight ω(t) through a piecewise linear function to adapt to the particle search capability adjustment under different perturbation levels: Where, δ θ (t) represents the disturbance intensity in the pitch direction during the t-th control cycle, δ φ (t) represents the disturbance intensity in the roll direction during the t-th control cycle, δ ψ (t) represents the disturbance intensity in the yaw angle direction during the t-th control cycle, where ω1, ω2 and ω3 are three preset inertial weight constants, and ω1>ω2>ω3, and ∈1 and ∈2 are the disturbance intensity segmentation threshold parameters; S34. Construct a fuzzy logic adjustment mechanism and set the attitude error change rate. With the local convergence amplitude ρ of the particle j As a fuzzy input, a self-learning factor c1(t) and a group learning factor c2(t) are set as fuzzy outputs, and the learning factors are adjusted through a fuzzy rule base. S35. By fusing the difference vectors between the current state of the particle and the individual optimal position and the group optimal position, the velocity and position are updated, guiding the particle to dynamically adjust its direction in the search space and evolve to the optimal control parameters: V j (t+1)=ω(t)·V j (t)+c1(t)·r1·(P j (t)-X j (t))+c2(t)·r2·(G(t)-X j (t)); X j (t+1)=X j (t)+V j (t+1); Among them, V j (t+1) represents the updated velocity vector of the j-th particle at the (t+1)-th iteration, V j (t) represents the updated velocity vector of the j-th particle in the t-th iteration, where r1 and r2 are random numbers, and P j X(t) represents the optimal position of the individual particle, G(t) represents the current global optimal position, and X... j (t+1) is the updated position vector of the j-th particle in the (t+1)-th iteration; S36. Evaluate the performance of the control parameter combinations obtained in each iteration using a multi-objective fitness function and update P. j When the maximum number of iterations or the convergence condition is met, the initial candidate solution set of the control parameters is output for G(t) and G(t).

6. The attitude stabilization control method for a marine rocket recovery platform based on particle swarm optimization according to claim 1, characterized in that, S4 specifically includes: S41. Initialize the candidate solution set {K1,K2,...,K} using the output control parameters. M } serves as the initial population for the Zebra Optimization Algorithm, where each control parameter is an individual m∈{1,2,...,M}, where M represents the total number of individuals controlling the parameter; S42. Construct a leadership scoring function L based on posture response data. m This is used to measure the overall ability of individual control parameters in terms of stability and perturbation adaptability, and to classify particle roles: Where λ is the weighting factor, STD(y) m (t) represents the platform attitude response sequence y. m The standard deviation of (t), T s,m Let Δy be the attitude convergence time corresponding to the m-th individual control parameter. m Let Δt be the magnitude of the platform attitude deviation caused by the m-th individual control parameter. m The disturbance recovery time required for the attitude deviation to recover from its maximum value to the steady-state range; S43. Assign all individuals according to the leadership rating function L. m Sort in descending order and select the top P percent of individuals by score as the leader set. The rest are as a group of following individuals. A differentiated search strategy is applied to each follower, and a location update operation is performed for each follower: K m (t+1)=K m (t)+α·rand()·(K best (t)-K m (t))+β·randn(); Among them, K m (t+1) is the updated position vector of the m-th control parameter individual after the (t+1)-th iteration, K m (t) is the updated position vector of the m-th control parameter individual after the t-th iteration, K best (t) represents the individual with the best control parameters in the current population, α is the zebra global tracking coefficient, β is the local perturbation coefficient, rand() represents a uniform random number in the interval [0,1], and randn() represents a normal distribution perturbation term; S44. Implement an escape update mechanism for the leader individual to avoid getting trapped in local optima, and perform the following escape update: K lead (t+1)=K lead (t)+γ·sign(K lead (t)-K mean (t))·D; Where γ is the escape intensity coefficient, K mean (t) represents the average control parameter vector of the current iteration population, D represents the preset maximum perturbation amplitude, sign() is the sign function, and K lead (t+1) represents the leader's control parameter vector after escape update, K lead (t) is the vector of control parameters for the leader at the t-th iteration; S45. Repeat steps S43 to S44 until the number of iterations reaches the set upper limit T. z Alternatively, after reaching the optimal fitness convergence threshold, output the final optimized control parameter set K. * .

7. The attitude stabilization control method for a marine rocket recovery platform based on particle swarm optimization according to claim 1, characterized in that, S5 specifically includes: S51. Establish a global search module for the solution space, mainly controlled by an improved particle swarm optimization algorithm, to generate multiple initial candidate solutions for control parameters and guide the overall evolution direction of the solution space. S52. Construct a local fine search module with the Zebra optimization algorithm as the slave control, and perform perturbation guidance and jump compensation on the candidate solutions generated by the master control module to enhance the local optimization ability and the ability to escape the local optimal trap. S53. Establish an information interaction mechanism between the master control and slave control modules, and periodically feed back the local optimal solution in the Zebra optimization algorithm to the particle swarm optimization algorithm to adjust the global guiding solution of the particle swarm and update the search direction. S54. Based on the performance of individual control parameters in the improved particle swarm optimization algorithm and zebra optimization algorithm, a multi-strategy joint scheduling module is constructed to dynamically determine whether the solution of each control parameter is suitable for global expansion or local convergence search path, and allocate it to the master control module and slave control module respectively for iterative update. S55. Construct a fusion optimization parameter evaluation mechanism, conduct a comprehensive fitness evaluation of the control parameter set output by each round of master-slave collaborative search, and extract the optimal control parameter set as the fusion optimization result. S56. By integrating the dynamic trends in the evaluation results, feedback is used to adjust the information exchange frequency and control parameter selection strategy between the master and slave modules, thereby achieving performance-driven collaborative optimization scheduling and structural adaptive adjustment.

8. The attitude stabilization control method for a marine rocket recovery platform based on particle swarm optimization according to claim 1, characterized in that, S6 specifically includes: S61. Input the optimized control parameter set after fusion and optimization into the fuzzy PID controller to configure the proportional, integral and derivative gain parameters of the three channels of the controller. S62. Read the real-time attitude input of the current platform from the attitude state time series dataset, including the state values ​​of pitch angle, roll angle and yaw angle, and use them as the input reference for the fuzzy PID controller. S63. The fuzzy PID controller calculates the control output of the three attitude channels, which correspond to the error signal adjustment results in the pitch, roll and yaw directions, respectively. S64. The control signal output by the controller is transmitted to the servo system in real time to drive the three-axis actuator to perform attitude adjustment in the pitch, roll and yaw directions respectively. The servo system is specifically composed of a three-axis electric actuator, a servo driver, a control signal receiving module and a feedback sensor module. It receives control commands, drives the platform attitude adjustment and provides real-time feedback on the execution status. S65. The servo system performs corresponding physical actions on the platform structure according to the received control commands, and completes the real-time correction and adjustment of the attitude. S66. After the platform attitude adjustment is completed, the latest response status data of each axis is collected to form the attitude response results, which are then fed back for performance evaluation and dynamic optimization.

Images