Supersonic civil aircraft control method and system based on multi-cell decomposition and readable storage medium

Abstract

The invention provides a supersonic civil aircraft control method and system based on multi-cell decomposition and a readable storage medium, which can be applied to control law design in a large envelope flight process of a supersonic civil aircraft, and comprises the following steps: S1, selecting scheduling parameters in a wide envelope of the aircraft and a change range thereof, performing high-density sampling in an envelope according to the scheduling parameters and the variation range thereof to generate a working point set, and acquiring a local linear model set at each working point by adopting a linearization method; s2, performing dimension reduction processing on the local linear model set by applying singular value decomposition, identifying a core parameter subspace and screening out a polyhedral vertex linear model forming a convex envelope; s3, aiming at each polyhedron vertex linear model, solving and designing a local state feedback controller through a linear matrix inequality by adopting a parallel distribution compensation framework; and S4, performing convex combination weighted fusion on the local state feedback controller by using a continuous weighting function, and generating a global control law to be applied to aircraft full envelope control.

Description

Technical Field

[0001] This invention relates to the field of supersonic civil aircraft control technology, and in particular to a supersonic civil aircraft control method, system, and readable storage medium based on multicellular decomposition. Background Technology

[0002] Currently, there are two main architectures for civil aircraft control: linear control and nonlinear control. Linear control methods have been studied earlier and have reached a relatively mature stage, with wider applications in engineering. PID control, gain scheduling, linear quadratic optimal control, pole placement, and tangent linearization have all achieved excellent control results in aircraft control research. Nonlinear control methods can further improve flight performance, with dynamic inverse control, feedback linearization, sliding mode control, and backstepping methods receiving extensive research in recent years. However, traditional linear control methods cannot adapt to large-scale parameter perturbations and must perform gain scheduling according to different parameters; while existing nonlinear control methods heavily rely on the object model, requiring large nonlinear gains for large disturbances and uncertainties, which can lead to chattering in the control system. Therefore, designing a control method with low model dependence and adaptability to large-scale parameter perturbations during large-envelope flight has become a research hotspot in supersonic civil aircraft control.

[0003] In patent CN120370812A, an equilibrium point is selected and the nonlinear model of the UAV is converted into a linear variable parameter (LPV) model through Jacobian linearization. A parameter-dependent state feedback controller is designed for the LPV model. A Lyapunov function suitable for the system is constructed, a set of linear matrix inequalities (LMI) conditions are derived, the control gain corresponding to the LMI conditions is obtained, gain scheduling is performed according to the parameters, and a gain scheduling controller that satisfies the system stability conditions is constructed.

[0004] The control strategy employed in supersonic civil aircraft has significant limitations when dealing with highly nonlinear systems. First, the selection of equilibrium points relies heavily on the designer's experience and intuition, lacking systematic guiding principles. This leads to subjectivity and arbitrariness, and makes it difficult to guarantee the completeness and representativeness of the selected points. Second, local controllers designed based on these discrete equilibrium points generally employ simple linear interpolation methods when extending to the full flight envelope. However, this interpolation strategy lacks rigorous mathematical foundations and cannot guarantee the stability and dynamic performance of the closed-loop system at non-design points within the envelope, posing a risk of control failure.

[0005] Therefore, it is necessary to study a control method, system and readable storage medium for supersonic civil aircraft based on multicellular decomposition to address the shortcomings of the existing technology and solve or mitigate one or more of the above problems. Summary of the Invention

[0006] In view of this, the present invention provides a supersonic civil aircraft control method, system and readable storage medium based on multicellular decomposition, which can be applied to the control law design during the large envelope flight process of supersonic civil aircraft.

[0007] On one hand, the present invention provides a supersonic civil aircraft control method based on multi-cell decomposition, the supersonic civil aircraft control method based on multi-cell decomposition includes the following steps: S1: Select the scheduling parameters and their variation range within the wide envelope of the aircraft, perform high-density sampling within the envelope based on the scheduling parameters and their variation range to generate a set of working points, and use a linearization method at each working point to obtain a set of local linear models; S2: Singular value decomposition is applied to the local linear model set to reduce dimensionality, identify the core parameter subspace, and select the polyhedral vertex linear models that constitute the convex envelope. The polyhedral vertex linear models cover the nonlinear dynamics of the entire envelope. S3: For the linear model of each polyhedron vertex, a local state feedback controller is designed by solving linear matrix inequalities using a parallel allocation compensation framework. S4: The local state feedback controller is subjected to convex combination weighted fusion using a continuous weight function to generate a global control law that is applied to the full envelope control of the aircraft.

[0008] As described above, and in any possible implementation, a further implementation is provided in S1: The scheduling parameters include Mach number, altitude, and angle of attack, and the range of variation covers all operating conditions within the wide envelope of the aircraft. The high-density sampling uses a uniform distribution method to generate dynamic datasets, and the sampling point spacing is set to be below a preset threshold. The linearization method is Jacobian linearization, which obtains a local linear model near the equilibrium point for each operating point; the set of local linear models forms a continuous trajectory in a high-dimensional parameter space, and the approximate accuracy is verified by comparing it with the nonlinear dynamics model of the aircraft.

[0009] As described above, and in any possible implementation, a further implementation is provided in S2: The singular value decomposition decomposes the dominant dynamic direction of the local linear model set and extracts the subspace corresponding to the dominant singular values. The subspace corresponding to the dominant singular values ​​is the core parameter subspace. The core parameter subspace is a reduced-dimensional space; the minimum vertex set is calculated using the convex hull algorithm within the core parameter subspace. The linear model of the polyhedron vertex is composed of the linear model corresponding to the minimum vertex set, and the convex hull precisely encloses the linear model corresponding to all sampled working points.

[0010] As described above, and in any possible implementation, a further implementation is provided in S3: The parallel allocation compensation framework transforms the global control problem into a set of independent linear matrix inequalities for each vertex. The linear matrix inequalities are constructed based on common Lyapunov functions; The state feedback gain of each vertex is obtained by solving linear matrix inequalities; the local state feedback controller is applied to the corresponding vertex closed-loop system.

[0011] As described above, and in any possible implementation, a further implementation is provided in S4: The continuous weighting function is a polyhedral membership function, which is associated with the position of the current working point within the convex hull. Calculate the weight values ​​for each vertex for the current working point; The local state feedback controller is weighted and summed according to the weight values.

[0012] In addition to the aspects described above and any possible implementation, a further implementation is provided in which: the dynamic dataset contains the aircraft state and parameters at all sampling working points; and the preset threshold is determined based on the nonlinear dynamic characteristics of the aircraft.

[0013] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein in S2: the convex hull algorithm traverses the dimensionality-reduced sampling point set; the minimum vertex set is the original working point corresponding to the boundary point of the convex hull.

[0014] In addition to the aspects described above and any possible implementation, a further implementation is provided in which: in S4: the weighted summation generates the global control gain corresponding to the current operating point; the global control law is obtained by multiplying the global control gain by the aircraft state.

[0015] In accordance with the aspects and any possible implementations described above, a supersonic civil aircraft control system based on multi-cell decomposition is further provided to implement the supersonic civil aircraft control method based on multi-cell decomposition, wherein the supersonic civil aircraft control method based on multi-cell decomposition includes the following steps: The point set model acquisition module is used to select the scheduling parameters and their variation range within the wide envelope of the aircraft, perform high-density sampling within the envelope based on the scheduling parameters and their variation range to generate a working point set, and use a linearization method at each working point to obtain a local linear model set; The dimension reduction module is used to apply singular value decomposition to the local linear model set for dimension reduction, identify the core parameter subspace and filter out the polyhedral vertex linear models that constitute the convex envelope, and the polyhedral vertex linear models cover the nonlinear dynamics of the entire envelope. The feedback control module is used to design a local state feedback controller for each polyhedron vertex linear model by using a parallel allocation compensation framework and solving linear matrix inequalities. The weighted fusion module is used to perform convex combination weighted fusion on the local state feedback controller using a continuous weight function to generate a global control law for application to the full envelope control of the aircraft.

[0016] In addition to the aspects described above and any possible implementations, a computer-readable storage medium is further provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the supersonic civil aircraft control method based on multicellular decomposition.

[0017] Compared with the prior art, the present invention can achieve the following technical effects: 1. Data-driven intelligent recognition and model conversion technology for multicellular vertices: Traditional methods rely on engineers' experience to select a limited number of equilibrium points within the flight envelope for linearization, a process that is subjective and incomplete. This invention proposes a data-driven automated modeling process. First, using a nonlinear mechanism model of a supersonic aircraft, high-density sampling is performed throughout the entire flight envelope (covering key parameters such as Mach number, altitude, and angle of attack) to generate a dynamic dataset covering the entire envelope. Then, dimensionality reduction techniques such as singular value decomposition or principal component analysis are applied to identify the core parameter directions that dominate the aircraft's dynamic characteristics. Finally, in these low-dimensional core parameter spaces, the convex hull algorithm is used to accurately calculate the smallest convex polyhedron that can enclose all sampled data points. The flight states corresponding to the vertices of this convex polyhedron constitute the "vertex system" required to construct the polytopic model. By performing Jacobian linearization at each vertex, a set of vertex linear models can be obtained, thereby accurately converting the original nonlinear model into a linear polytopic model. 2. Conservative optimization techniques for vertex models based on parameter-dependent Lyapunov functions: The conservatism of multi-cell models mainly stems from the fact that using a convex combination of a finite number of vertex systems to describe an infinite number of interior point systems introduces uncertainty, leading to overly conservative controller designs and poor performance. This invention aims to reduce this conservatism at its root. The core idea is not to directly optimize the model itself, but rather to optimize the "metric" describing the model's dynamic stability. This involves introducing a parametrically dependent Lyapunov function (LMI), whose functional form (e.g., matrix elements) changes continuously with flight parameters (e.g., Mach number, altitude), rather than the traditional fixed LMI. In the controller design phase, this invention takes "finding a parametrically dependent Lyapunov function that can stabilize all vertex systems" as the optimization objective, transforming it into a feasibility problem involving a series of linear matrix inequalities. By solving the LMI, this invention obtains not only a controller but also a stability criterion that closely reflects the system's true dynamic changes. This process is equivalent to finding the "tightest" convex envelope describing the system's dynamics while ensuring stability, thereby minimizing the conservatism caused by model approximation. 3. Vertex controller synthesis and full envelope stability assurance technology based on convex combination theory: After obtaining a low-conservatism multicell model, this invention proposes a systematic framework for controller design and stability proof. First, for each optimized vertex system, a vertex local state feedback controller is designed using LMI technology, ensuring that this controller and the corresponding parameter-dependent Lyapunov function jointly satisfy the stability and performance indicators of the vertex system. The most crucial step lies in the theoretical guarantee of the stability of the entire flight envelope: using convex combination theory as the core tool, any interior point within the flight envelope can be represented as a convex combination of vertex systems. The global controller designed in this invention also adopts the same convex combination method, i.e., it is obtained by weighted summation of the vertex controllers according to the parameter weights of the interior point. According to the properties of convex combination theory, if all vertex systems are stable under their respective controllers, then any interior point closed-loop system formed by convex combinations of these vertex controllers and vertex systems with the same weights must also be stable. This provides a rigorous, non-conservative mathematical proof for the stability of the controller throughout the entire continuous flight envelope.

[0018] Of course, any product implementing this invention does not necessarily need to achieve all of the technical effects described above at the same time. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the design of a supersonic civil aircraft control method based on multi-cell decomposition, provided in one embodiment of the present invention. Figure 2 This is a diagram illustrating the construction of a multi-cell model of a supersonic civil aircraft according to an embodiment of the present invention; Figure 3 This is a PDC control architecture diagram provided in one embodiment of the present invention; Figure 4 This is a diagram illustrating the control effect of a supersonic civil aircraft based on a multicellular architecture, provided by an embodiment of the present invention. Detailed Implementation

[0021] To better understand the technical solution of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0023] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0024] To address the wide speed range, large flight envelope, and strong coupling characteristics of supersonic civil aircraft, this invention proposes a modeling and control synthesis method based on polycellular decomposition. This method innovatively transforms the full-envelope nonlinear dynamic model of the civil aircraft into a convex envelope composed of linear models of the vertices of polycellular bodies. The core advantage of this method lies in achieving coverage of the nonlinear system through polycellular decomposition and extending the design and verification of control laws from finite equilibrium points to the entire continuous flight envelope using convex combination theory, providing a systematic and highly robust solution to the challenge of full-envelope control of strongly nonlinear systems. This invention provides a supersonic civil aircraft control method based on polycellular decomposition. The method includes the following steps.

[0025] Step S1: Select the scheduling parameters and their variation range within the wide envelope of the aircraft. Based on the scheduling parameters and their variation range, perform high-density sampling within the envelope to generate a set of working points. At each working point, use a linearization method to obtain a set of local linear models.

[0026] Specifically, the wide envelope of an aircraft typically refers to the range of all flight conditions that an aircraft may encounter throughout its entire flight mission, including its dynamic characteristics under different altitudes, speeds, and attitudes. Scheduling parameters are key variables that characterize the aircraft's current flight state and significantly influence its dynamic characteristics. In one embodiment, scheduling parameters include Mach number, altitude, and angle of attack. These three parameters comprehensively cover changes in various typical operating conditions, such as low-speed to high-speed, low-altitude to high-altitude, and level flight to high angle-of-attack maneuvers.

[0027] The range of variation for scheduling parameters needs to be determined based on the specific design envelope of the aircraft. For example, Mach number can range from 0.3 to 2.5, altitude from sea level to 20 kilometers, and angle of attack from -5 degrees to 25 degrees to ensure that all possible flight conditions are included. It should be noted that the determination of the variation range should be based on the mission requirements and aerodynamic data of the aircraft to ensure that the envelope boundaries can accommodate extreme operating conditions.

[0028] After determining the scheduling parameters and their range of variation, high-density sampling is performed within the envelope to generate a set of working points. The purpose of high-density sampling is to capture the changing trends of the aircraft's nonlinear dynamics as densely as possible, avoiding the omission of key dynamic regions. In one implementation, a uniform distribution sampling method is used, that is, the grid is divided at equal intervals along each scheduling parameter dimension. For example, 200 points are divided in the Mach number direction, 100 points in the altitude direction, and 50 points in the angle of attack direction, thereby generating a set of more than 1 million sampling working points. The sampling point spacing is set below a preset threshold, which can be determined according to the degree of nonlinearity of the aircraft. For example, the Mach number spacing does not exceed 0.01, the altitude spacing does not exceed 200 meters, and the angle of attack spacing does not exceed 0.5 degrees, to ensure the effectiveness of the local linearization approximation.

[0029] At each sampling operating point, a local linear model is obtained using a linearization method. Specifically, the linearization method is Jacobian linearization, which involves performing a first-order Taylor expansion of the nonlinear dynamic equations of the aircraft near the equilibrium point corresponding to that operating point to obtain a local linear model in state-space form. This local linear model is typically represented by a state matrix A, a control matrix B, and the corresponding output matrix. The equilibrium point is obtained through balancing calculations, i.e., finding the steady-state solution that makes the derivative of the nonlinear equations zero. For each operating point, balancing and linearization are performed independently, thus forming a set of local linear models covering the entire envelope.

[0030] For example, for a certain type of supersonic vehicle, at the operating point of Mach 1.6, altitude 15 km, and angle of attack 8 degrees, the aerodynamic coefficients and mass characteristics at that point are first calculated to determine the thrust-drag matching and lift-gravity matching under equilibrium conditions. Then, at this equilibrium point, the state-to-state Jacobian matrix and the control-to-state Jacobian matrix are calculated to obtain the local linear model for that point. This process is repeated for all sampling points, ultimately yielding a set of local linear models in a high-dimensional parameter space.

[0031] The local linear model set forms a continuous trajectory in a high-dimensional parameter space, reflecting the dynamic evolution of the aircraft as scheduling parameters change. The approximation accuracy is verified by comparing it with the original nonlinear dynamic model. For example, by applying small perturbations near each operating point, the consistency between the linear model's predicted response and the nonlinear simulation response is compared to ensure that the linearization error is within an acceptable range. This verification process helps confirm whether the sampling density is sufficient; if the error is large, the sampling can be further densified.

[0032] Step S2: Apply singular value decomposition to the set of local linear models for dimensionality reduction, identify the core parameter subspace, and select the polyhedral vertex linear models that constitute the convex envelope. The polyhedral vertex linear models cover the nonlinear dynamics of the entire envelope.

[0033] Specifically, the local linear model set contains a large number of state and control matrices corresponding to operating points, and these matrix elements form a high-dimensional data cloud as scheduling parameters change. Designing the controller directly on the entire model would face the curse of dimensionality and excessive computational complexity. Therefore, dimensionality reduction is necessary to extract the dominant dynamic information.

[0034] In one implementation, the state matrix A of the local linear model set is reorganized, arranging the matrix elements of all sampling points into a high-dimensional vector to form a data matrix. Then, singular value decomposition (SVD) is applied to this data matrix. SVD can decompose the dominant dynamic direction; that is, the singular values ​​are arranged from largest to smallest, and the left and right singular vectors corresponding to the first few large singular values ​​represent the main patterns of model change.

[0035] The subspace corresponding to the dominant singular values ​​is the core parameter subspace. Specifically, by setting a singular value decay threshold, such as retaining the dimension corresponding to singular values ​​with accumulated energy of 99% or more, the original high-dimensional space can typically be reduced to a low-dimensional space of 3 to 6 dimensions. This core parameter subspace retains the most important characteristics of the aircraft's dynamic changes, such as the changing trends of dominant modes like long-period and short-period modes, while filtering out noise and secondary disturbances.

[0036] Within the reduced core parameter subspace, the linear models corresponding to all sampled operating points are projected onto this lower-dimensional space, forming a low-dimensional point set. Then, the convex hull algorithm is applied to compute the minimum vertex set. The convex hull algorithm traverses the reduced sampled point set, searching for extrema that constitute the outer boundary; these extrema correspond to specific operating points in the original high-dimensional space.

[0037] For example, assuming the core parameter subspace is 4-dimensional after dimensionality reduction, all sampled points are projected as a 4-dimensional point cloud. The convex hull algorithm first identifies boundary points, such as using the Quickhull algorithm or Gift Wrapping algorithm, and gradually constructs the boundary of the convex polyhedron starting from the outermost points, ultimately obtaining the minimum number of vertices. The original working points corresponding to these vertices are the linear model of the polyhedron vertices.

[0038] The polyhedral vertex linear model is composed of the linear models corresponding to the minimum set of vertices. Typically, the number of vertices is much smaller than the original sampling points, for example, reduced from hundreds of thousands of sampling points to a few dozen vertex models. These vertex models are located at dynamically changing extreme positions and can accurately enclose the linear models corresponding to all sampling points through convex combinations, thereby covering the nonlinear dynamics of the entire envelope.

[0039] It should be noted that the accuracy of the convex hull is confirmed by verifying that all projected points are inside the convex hull. If any sampling points fall outside the convex hull, the dimensionality reduction needs to be adjusted or the sampling density increased. This step significantly reduces the computational burden of subsequent controller design while ensuring global coverage.

[0040] In one possible implementation, for longitudinal channel control, singular value decomposition retains the first three dominant directions, corresponding to the main changes in short-period, long-period, and height modes. Convex hull calculation yields 12 vertex models, corresponding to extreme conditions such as maximum Mach number, minimum Mach number, highest altitude, lowest altitude, and maximum angle of attack at the envelope boundaries. Through convex combinations of these vertex models, the dynamic characteristics of any operating point within the envelope can be approximated.

[0041] Step S3: For each polyhedron vertex linear model, a local state feedback controller is designed by solving linear matrix inequalities using a parallel allocation compensation framework.

[0042] Specifically, the number of linear models of polyhedron vertices is limited, typically between 10 and 50, making it feasible to design an independent local state feedback controller for each vertex. The parallel allocation compensation framework is an efficient control design method that transforms the global nonlinear control problem into multiple independent vertex linear control subproblems while ensuring overall stability.

[0043] In one embodiment, the parallel allocation compensation framework constructs a set of linear matrix inequalities based on a common Lyapunov function. The common Lyapunov function ensures that all vertex closed-loop systems share the same quadratic positive definite function, thereby providing a stability guarantee for subsequent convex combinatorial fusion.

[0044] For each vertex linear model, the local state feedback controller takes the form of the product of the state feedback gain and the state. Design objectives include pole placement, robust stability, and disturbance suppression, which are typically transformed into linear matrix inequality constraints. For example, it may require the eigenvalues ​​of the closed-loop matrix to lie in a specific region of the left half-plane, or to satisfy the performance index of H infinity.

[0045] A linear matrix inequality solver is used to simultaneously solve the same set of linear matrix inequality variables for all vertex models, including the common positive definite matrix and the feedback gain matrix of each vertex. This parallel solution method fully utilizes the independence of the vertex models and can significantly improve computational efficiency.

[0046] For example, for a certain vertex model corresponding to a hypersonic high angle of attack condition, its local linear model exhibits strong short-period instability. By adding pole constraints through linear matrix inequalities, the short-period modes are configured to the region with a damping ratio of 0.7 and a natural frequency of 15 rad / s. At the same time, the same Lyapunov matrix is ​​shared for all vertices to ensure controllable conservatism.

[0047] After the local state feedback controller is applied to the corresponding vertex closed-loop system, each vertex achieves local asymptotic stability. Due to the use of a common Lyapunov function, the stability of the global closed-loop system can be directly inherited in subsequent convex combinations. This design avoids the excessive conservatism that might result from point-by-point independent Lyapunov functions.

[0048] Step S4: The local state feedback controller is subjected to convex combination weighted fusion using a continuous weight function to generate a global control law that is applied to the full envelope control of the aircraft.

[0049] Specifically, in the preceding steps, local state feedback controllers were designed for each polyhedral vertex linear model. These controllers only exhibit optimal performance at their corresponding vertices, while the actual operating point of the aircraft typically lies at any location within the convex envelope. Therefore, a convex combination weighted fusion method is needed to smoothly transition multiple local controllers into a single global control law to adapt to dynamic changes within the entire envelope. Continuous weighting functions are the key tool for achieving this fusion; they dynamically allocate the contribution ratio of each vertex controller based on the current operating point's position within the convex envelope.

[0050] In one implementation, the continuous weighting function is designed as a polyhedral membership function. This function characterizes the geometric relationship between the current working point and each vertex, typically calculated based on the projection position of the scheduling parameters within the core parameter subspace. The weighting function satisfies the convex combination property, meaning the sum of all weight values ​​is 1, and each weight value varies between 0 and 1, ensuring the continuity and stability of the fusion result.

[0051] Step S51: The continuous weighting function is a polyhedral membership function, which is associated with the position of the current working point within the convex hull.

[0052] Specifically, the polyhedral membership function determines the weight values ​​by calculating the relative distances or geometric projections from the current working point to each vertex. One possible implementation uses a barycentric coordinate-based method, representing the current working point as a linear combination of vertices within a convex envelope, with the barycentric coordinate coefficients representing the corresponding weight values. This method ensures that the weight function changes continuously throughout the envelope, avoiding control smoothness issues caused by abrupt changes.

[0053] For example, for a tetrahedral convex hull consisting of four vertices, the position of the current working point within the core parameter subspace can be determined by the ratio of its distances to each vertex. If the working point is close to a certain vertex, the weight value corresponding to that vertex is larger, close to 1, while the weight values ​​of other vertices are smaller. This weight allocation method makes the control law more inclined to adopt the controller characteristics of vertices closer to the working point, thereby improving local adaptability.

[0054] It should be noted that the design of the membership function can also consider the physical meaning of the scheduling parameters. For example, in aircraft control, if the current Mach number is close to the Mach number corresponding to a certain vertex, the weight value of that vertex can be appropriately increased to better match the dynamic characteristics under high-speed or low-speed conditions. This adjustment based on physical meaning can further improve the adaptability of the control law to nonlinear dynamics.

[0055] Step S52: Calculate the weight values ​​corresponding to each vertex for the current working point.

[0056] In one embodiment, the weight calculation process first obtains the scheduling parameter values ​​of the aircraft's current state, such as Mach number, altitude, and angle of attack measured in real time by sensors. Then, these parameter values ​​are projected into the core parameter subspace to obtain the coordinates of the current operating point in a low-dimensional space. Based on the polyhedral membership function, the geometric relationships from this position to each vertex are calculated to generate the corresponding weight values.

[0057] For example, suppose the current operating point corresponds to a Mach number of 1.2, an altitude of 10 kilometers, and an angle of attack of 5 degrees. After projection, it is located in the middle position near the two vertices in the core parameter subspace. Using the barycentric coordinates, the weights of these two vertices are calculated to be 0.6 and 0.3, respectively, while the weights of the other vertices are smaller, at 0.05 and 0.05. This weight allocation reflects the dynamic characteristic that the current operating point is closer to the first two vertices.

[0058] In another implementation, the weight calculation can also incorporate a smoothing filtering mechanism to avoid frequent jumps in weight values ​​caused by sensor noise or measurement errors. For example, averaging the weight values ​​over a continuous time period ensures the smoothness of control law changes. This method is particularly suitable for the control requirements of high-speed aircraft in turbulent environments.

[0059] Step S53: Perform a weighted summation on the local state feedback controller according to the weight values.

[0060] Specifically, after obtaining the weight values ​​of each vertex, these weight values ​​are weighted and summed with the state feedback gain matrix of the corresponding vertex to generate the global control gain corresponding to the current operating point. The global control gain is a linear combination of multiple local controllers, inheriting the stability and performance characteristics of each vertex controller.

[0061] In one possible implementation, the weighted summation process is performed on each element of the state feedback gain matrix separately, ensuring that the generated global control gain matrix transitions smoothly. For example, if the gain matrix of a vertex controller has a large feedback coefficient for pitch velocity, and the current operating point is close to that vertex, the feedback coefficient for pitch velocity in the global control gain will also increase accordingly, thereby enhancing the control capability for short-period modes.

[0062] Step S81: The weighted summation generates the global control gain corresponding to the current operating point.

[0063] Specifically, the global control gain is generated by linearly superimposing the state feedback gain matrices of each vertex according to their weights to form a new gain matrix. This gain matrix can dynamically adapt to the nonlinear characteristics of the current operating point while preserving the local optimization performance of each vertex controller.

[0064] For example, assuming the current operating point is close to the high-speed, high-angle-of-attack vertex with a weight of 0.7, while the weight is close to the low-speed, small-angle-of-attack vertex with a weight of 0.3, then in the generated global control gain matrix, the gain characteristic of the high-speed, high-angle-of-attack vertex controller dominates. This dynamic adjustment method allows the control law to exhibit targeted control effects in different regions within the control envelope.

[0065] It should be noted that the generation process of the global control gain can be performed in real time, meaning that the weight values ​​and global control gain are continuously updated as the aircraft state changes. This real-time capability is particularly important for dealing with rapidly changing flight conditions, especially in supersonic or high-maneuver flight, effectively avoiding the instability risks caused by control hysteresis.

[0066] Step S82, the global control law is obtained by multiplying the global control gain by the aircraft state.

[0067] In one embodiment, the global control law is calculated by multiplying the global control gain matrix by the current aircraft state vector to obtain the control input. The aircraft state vector typically includes key state quantities such as angular velocity, attitude angle, and velocity components, which are acquired in real time by sensors. The control input corresponds to the aircraft's actuator commands, such as control surface deflection angle and thrust magnitude.

[0068] For example, for a certain type of aircraft, the current state vector includes pitch rate, roll rate, and yaw rate, and the global control gain matrix contains feedback coefficients for these state variables. Through matrix multiplication, the deflection commands corresponding to the elevator, rudder, and ailerons are calculated. These commands are directly sent to the actuators to achieve closed-loop control of the aircraft.

[0069] In another implementation, the global control law can also incorporate saturation and rate limits to prevent actuator overload or excessively rapid response. For example, upper and lower limits can be set for the control surface deflection command to ensure it remains within mechanical limits, while the rate of command change can be limited to prevent structural vibration caused by rapid changes. This protection mechanism enhances the reliability of the control law in practical applications.

[0070] In one possible implementation, the design of the global control law can also consider the needs of different flight phases. For example, during takeoff and landing, the control law focuses more on low-speed stability, which can be achieved by adjusting the weighting function and increasing the weight value of the low-speed peak controller; while during cruise, fuel economy is prioritized, and the amplitude of the control input can be appropriately reduced. This phased adjustment approach allows the control law to better adapt to mission requirements.

[0071] It should be noted that the continuity of the global control law benefits from the smoothness of the weighting function, avoiding the abrupt switching problems that may occur with traditional piecewise control laws. This smooth transition characteristic is particularly important for scenarios where aircraft frequently change operating conditions within a wide control envelope, significantly improving flight stability and passenger comfort.

[0072] In one embodiment, for a supersonic vehicle accelerating from subsonic to supersonic speeds, the control law needs to adapt to the drastic changes in aerodynamic characteristics. By updating weight values ​​in real time, a gradual transition from a subsonic vertex controller to a supersonic vertex controller is achieved, ensuring the continuity and stability of the control input throughout the acceleration process. This dynamic adaptability effectively addresses the nonlinear challenges across the supersonic region.

[0073] In another implementation, the global control law can also be implemented in conjunction with the aircraft's mission configuration file. For example, during high-maneuverability missions, a preset weight adjustment strategy can prioritize the characteristics of the high angle-of-attack vertex controller to enhance maneuverability; while during long-endurance missions, the characteristics of the cruise vertex controller are prioritized to optimize energy consumption. This mission-oriented weight adjustment method can further improve the applicability of the control law.

[0074] Specifically, the design of the weighting function can also incorporate time-dependent factors, that is, consider the historical trajectory of changes in the aircraft's state. For example, by analyzing the changing trends of scheduling parameters over a period of time, the dynamic evolution direction of the current operating point can be predicted, and the weight values ​​can be adjusted in advance. This predictive adjustment can reduce the overreaction of the control law to instantaneous disturbances and improve the overall control quality.

[0075] For example, during the descent of an aircraft from high altitude and high speed to low altitude and low speed, the weighting function can preemptively increase the weight value of the low-speed, low-altitude apex controller based on historical trajectories, ensuring that the control law completes most of the transition before entering the target operating condition. This proactive adjustment effectively avoids control mismatch problems caused by rapid changes in operating conditions.

[0076] In one possible implementation, the global control law can be verified using a high-fidelity simulation environment. This environment includes a nonlinear dynamics model of the aircraft, an aerodynamic database, and an environmental disturbance model. The stability, robustness, and performance metrics of the control law are evaluated by testing its response characteristics under different flight conditions. For example, gust disturbances can be simulated in the simulation to observe whether the control law can quickly stabilize the aircraft's attitude.

[0077] It should be noted that simulation verification can also cover boundary conditions, such as extreme scenarios like maximum Mach number, maximum altitude, and maximum angle of attack, ensuring that the control law remains effective in these regions. This comprehensive verification allows for the identification of potential deficiencies in the weighting function or global control gain, enabling further optimization of design parameters.

[0078] In one embodiment, for a certain type of high-maneuverability aircraft, a high angle-of-attack dive maneuver is simulated in the simulation to test the performance of the global control law under extreme conditions. By adjusting the smoothing parameters of the weighting function, it is ensured that the control input does not undergo abrupt changes due to vertex switching, while simultaneously verifying the control law's adaptability to aerodynamic parameter uncertainties. This verification method can provide a reliable basis for the practical application of the control law.

[0079] In another implementation, the global control law can be combined with an online learning mechanism. For example, during flight, by collecting actual flight data, the parameters of the weighting function are dynamically updated to adapt to changes in the aircraft's dynamic characteristics caused by changes in fuel consumption or payload. This adaptive capability can further improve the performance of the control law in long-endurance missions.

[0080] Specifically, the online learning mechanism can analyze whether the current weighting function accurately matches the actual dynamic characteristics by recording the relationship between the aircraft's state response and the control input. If a large deviation is found, the parameters of the weighting function can be fine-tuned using gradient descent or other optimization methods. This fine-tuning process usually runs in the background and does not affect the execution of real-time control.

[0081] For example, during long-endurance flight missions, fuel consumption causes the aircraft's center of gravity to shift rearward, altering its dynamic characteristics. Through an online learning mechanism, the weight function is gradually adjusted, increasing the weight value of the vertex controller corresponding to the rearward shift of the center of gravity, thereby ensuring that the control law always maintains an optimal match with the current dynamic characteristics.

[0082] In one possible implementation, the design of the global control law can also consider multi-channel coupling effects. For example, there are coupling relationships between the longitudinal, lateral, and directional channels of an aircraft, and a control law designed solely based on the longitudinal vertex may not be fully adaptable to lateral disturbances. Therefore, a channel coupling factor can be introduced into the weighting function to dynamically adjust the distribution ratio of control gains for each channel.

[0083] It should be noted that the channel coupling factor can be determined through offline simulation or flight test data analysis. For example, during high angle-of-attack flight, lateral roll has a significant impact on longitudinal pitch, so the weighting value of the lateral channel control gain can be appropriately increased. This multi-channel cooperative design can improve the overall performance of the control law in complex maneuvers.

[0084] In one embodiment, for a certain type of supersonic fighter jet, when performing a high angle-of-attack roll maneuver, a channel coupling factor is introduced, and the weighting function simultaneously considers the characteristics of the longitudinal and lateral vertex controllers to generate a comprehensive control gain. This comprehensive control method can effectively suppress the instability effects caused by inter-channel coupling and ensure attitude controllability during the maneuver.

[0085] In another implementation, the global control law can also incorporate the nonlinear characteristics of the aircraft's actuators. For example, if the control surface deflection exhibits nonlinear saturation or hysteresis, a nonlinear compensation stage can be added after the control law output. Through inverse mapping or feedback correction, the impact of actuator nonlinearity on the control effect can be offset.

[0086] Specifically, the design of the nonlinear compensation stage is based on the physical model of the actuator, such as the relationship between the steering surface deflection angle and the hydraulic system response. By adding a compensation function between the control law output and the actuator input, it is ensured that the actual control input is as consistent as possible with the desired control input. This compensation mechanism can improve the execution accuracy of the control law under highly dynamic conditions.

[0087] For example, during high-speed flight, the control surface deflection angle is significantly affected by aerodynamic loads, resulting in response hysteresis. Through a nonlinear compensation mechanism, the control law output anticipates the hysteresis and adjusts the command amplitude and timing to ensure that the actual control surface deflection angle matches the desired value. This compensation method can significantly improve control response speed and accuracy.

[0088] In one possible implementation, the deployment of the global control law can also take into account hardware implementation constraints. For example, in a flight control computer, the calculation of the control law needs to meet real-time requirements. This can be achieved by simplifying the calculation of the weight function or pre-storing the vertex control gain matrix, thereby reducing the amount of online computation. This optimization approach ensures the feasibility of the control law in resource-constrained environments.

[0089] It should be noted that hardware implementation constraints also include the digital processing of the control law's output signals. For example, control input commands need to be converted into discrete signals and sent to the actuator. This can be achieved through appropriate sampling frequency and quantization precision design, avoiding the impact of signal distortion on control performance. Such detailed optimization can further improve the engineering applicability of the control law.

[0090] In one embodiment, for a certain type of UAV platform, the control law is deployed in an embedded control unit. By pre-storing the vertex control gain matrix and simplifying the weight calculation method, the control law update cycle is controlled to within 5 milliseconds. This efficient implementation can meet the high-frequency control requirements of UAVs, ensuring real-time performance and reliability.

[0091] In another implementation, the design of the global control law can also consider fault tolerance. For example, when some sensors or actuators fail, a fault-tolerant control law can be generated by adjusting the weighting function to prioritize the characteristics of unaffected vertex controllers. This fault-tolerant design can improve the aircraft's survivability in abnormal conditions.

[0092] Specifically, the fault tolerance mechanism can be implemented through an online diagnostic module, such as by checking the consistency of sensor data to identify fault states. Subsequently, the weight function masks the weight values ​​of the vertex controllers associated with the fault and increases the weight ratio of other vertex controllers to ensure that the control law can still maintain basic control functions.

[0093] For example, when the altitude sensor fails, the weighting function can mask the vertex controller weights that depend on the altitude parameter, prioritizing the use of vertex controller characteristics based on Mach number and angle of attack to generate a fault-tolerant control law. This fault-tolerant approach can maintain the aircraft's basic attitude control capabilities even in the event of sensor failure.

[0094] In one possible implementation, the optimization of the global control law can also incorporate the uncertainties of the aircraft's aerodynamic parameters. For example, aerodynamic coefficients may be affected by environmental changes or manufacturing deviations. By introducing a robustness factor into the weighting function, the conservatism of the control gain can be dynamically adjusted to accommodate the effects of these uncertainties.

[0095] It should be noted that the robustness factor can be determined through Monte Carlo simulation, that is, by testing the control law performance under different aerodynamic parameter disturbances to find the optimal parameter range of the weighting function. This robust design ensures the control law's adaptability to parameter uncertainties in actual flight.

[0096] In one embodiment, for a certain type of high-altitude long-endurance aircraft, where aerodynamic parameters are significantly affected by the thin atmosphere at high altitudes, a robustness factor is introduced. This factor increases the conservative control gain of the weighting function under high-altitude conditions, ensuring that the control law maintains stability even with large deviations in aerodynamic parameters. This design significantly improves the reliability of the aircraft in extreme environments.

[0097] In another implementation, the global control law can also be combined with the multi-objective optimization requirements of the aircraft. For example, when designing the weight function, multiple objectives such as stability, maneuverability, and energy consumption can be considered simultaneously. Through multi-objective optimization methods, the weight values ​​of each vertex controller can be dynamically adjusted to achieve optimal overall performance.

[0098] Specifically, multi-objective optimization can be achieved through weighted summation or Pareto front methods. For example, a higher weight can be assigned to the stability objective, while a lower weight can be assigned to the energy consumption objective, generating a comprehensive weight function. This multi-objective design enables the control law to achieve balanced optimization under different task requirements.

[0099] For example, when performing long-range cruise missions, the weighting function can reduce the weight value of the maneuverability-related vertex controller and increase the weight value of the stability-related vertex controller, while reducing energy consumption caused by thrust changes by limiting the control input amplitude. This optimization method can effectively extend the aircraft's endurance.

[0100] In one possible implementation, the testing of the global control law can also be combined with actual flight tests. For example, in flight tests, by recording the response data of the control law under different operating conditions, the actual effects of the weighting function and the global control gain can be analyzed to verify its adaptability to nonlinear dynamics. This experimental verification can provide an important basis for the finalization of the control law.

[0101] It should be noted that flight tests can cover a variety of mission scenarios, such as takeoff, cruise, maneuvering, and landing, to ensure the applicability of the control law throughout the entire mission cycle. Test data can also be used to further optimize the parameters of the weighting function, improving the engineering sophistication of the control law.

[0102] In one embodiment, for a certain type of supersonic transport aircraft, the performance of the control law in the transonic region is tested during flight tests. By recording the control input and the aircraft's state response, the smoothness of the weighting function under drastic Mach number changes is analyzed. The test results are used to adjust the smoothing parameters of the weighting function to ensure the stability of the control law during transonic transitions.

[0103] In another implementation, the design of the global control law can also consider integration with other flight control modules. For example, when integrated with an autopilot or flight management system, the control law needs to provide standardized input / output interfaces while supporting dynamic adjustments to external commands. This integrated design can improve the compatibility of the control law in complex flight systems.

[0104] Specifically, the integration process can be achieved by defining standardized state vectors and control input formats. For example, the output commands of the global control law can be formatted as standard control surface deflection and thrust commands, seamlessly integrating with the autopilot's control logic. This standardized design simplifies the porting of control laws across different flight platforms.

[0105] Example 1: This invention addresses multiple technical challenges faced by the control systems of supersonic civil aircraft in complex flight environments, including difficulties in modeling strongly nonlinear dynamics, insufficient control accuracy under multi-physics coupling effects, and poor system robustness under drastic changes in flight conditions. Furthermore, this invention proposes and systematizes for the first time a "partitioned control architecture based on multi-cell decomposition," solving the problem that traditional linear control methods struggle to balance global performance and local stability across the wide envelope of supersonic flight. This method also reveals the potential optimization space for the control architecture in modular design, fault diagnosis, and fault-tolerant control, providing an innovative solution for designing high-performance, high-reliability supersonic civil aircraft control systems. The control law design process is as follows: Figure 1 As shown, First, establish a multicellular model of a supersonic civilian aircraft, such as Figure 2 As shown: The first step is to select appropriate scheduling variables and their range of variation based on the operating conditions and flight status of the supersonic civil aircraft; The second step is to compare the advantages and disadvantages of different linearization methods, select a mature and reliable linearization method, and obtain the aircraft linearization model near the equilibrium point based on the selected scheduling variables. This invention intends to use a dense point selection method to make up for the defect of arbitrary selection of equilibrium points in traditional linearization models, thereby reducing the error of the LPV model. The third step uses higher-order singular value decomposition theory to obtain the core subsystems in the LPV model and constructs a continuous weight function. This reduces the workload of control law design while establishing a mapping relationship between the subsystems and the entire flight process. The established multi-cell model is then compared with the nonlinear model of the civil aircraft to complete the approximation analysis.

[0106] Secondly, the control law was designed for the multi-cell model of the supersonic civil aircraft.

[0107] The first step employs a parallel allocation compensation framework, transforming the control problem into solving a series of linear matrix inequalities. This design process utilizes the LMI optimization tool to ensure that each vertex closed-loop subsystem meets preset stability and performance targets. The second step involves convex combination weighting of the individual local controllers using the same weighting function (or membership function) as the system model. This process seamlessly extends and merges the discrete vertex controllers into a global, smooth nonlinear controller.

[0108] In summary, this invention solves the problems of "blindness in equilibrium point selection and reliance on experience in controller design." The multicellular model, through systematic vertex selection, discretizes the continuously changing flight envelope into a finite, strictly convex geometric structure. This allows control design to no longer rely on randomly selected equilibrium points, but instead focus on a few decisive vertex systems, fundamentally eliminating subjective arbitrariness in the design process. Secondly, it solves the problem of "lack of theoretical guarantees for the performance and stability of the full-coverage controller." For example... Figure 3 As shown, the PDC framework constructs a global controller by convexly combining vertex controllers through weight functions. This structure naturally inherits the convexity of the multicellular model, allowing for rigorous mathematical proofs of the stability of the closed-loop system within the entire flight envelope using powerful theoretical tools such as Lyapunov functions, rather than just simulation verification. This provides a solid theoretical foundation for the safe control of supersonic civil aircraft under extreme and dynamically changing flight conditions.

[0109] First, based on the longitudinal nonlinear model of the supersonic civil aircraft, the state variable is selected as velocity. Angle of attack pitch angular velocity Pitch angle and height The control input is , representing the thrust and elevator deflection, respectively.

[0110] (1) LPV modeling is performed using the Jacobian linearization method, selecting flight speed and altitude as scheduling variables for the flight process. To avoid randomness in point selection during linearization, a dense point selection method is used, and for each... Find the equilibrium point, set the derivatives of all state variables to 0, and calculate the Jacobian matrix around the selected equilibrium point. Then the Jacobian matrix... , They are respectively: (2) (3) in: (4) To simplify the obtained equilibrium operating points, a high-order singular value decomposition method is used to screen the core operating points and construct the weight function for the longitudinal model of the supersonic aircraft, thus obtaining a polytopic model of the aircraft to accurately describe the complete flight envelope. Furthermore, to reduce the complexity of subsequent controller design, the weight function is further processed to obtain a standardized polytopic model that satisfies a strictly convex hull. (Supersonic Civil Aircraft Polytopic Model) As shown in equation (5), the scheduling variable , The tensor product modeling process is as follows: (5) Next, the tensor along Expanding each order separately yields matrices of different orders. Then, SVD decomposition is performed on the expanded matrices of each order. An SVD decomposition method based on Jacobi orthogonality is adopted, which achieves overall matrix orthogonality by orthogonalizing each column of the matrix. Specifically, this involves the following three steps: First, Algorithm 1 is used to solve for the orthogonal rotation matrix. Implement matrix and Orthogonalization of the columns will yield rotation matrices in sequence. Multiplication yields a right singular matrix. Next, calculate the orthogonalization matrix. Singular Matrix , of which elements The final left singular matrix is: The singular value decomposition result based on Jacobi orthogonalization is: .

[0111] Algorithm 1 Jacobian Rotation Matrix Calculation method enter: , Output: Jacobian orthogonal rotation matrix 1. ; 2. ; 3. ; 4. ; 5. ; After the matrices are expanded to each order, the left singular value matrix is ​​obtained by SVD decomposition. Substitute the values ​​and calculate the tensor. core tensor With approximate tensor .

[0112] (6) Among them, the approximate tensor It is Tensors of order, By vertex system constitute, The weight function corresponding to the vertex system is given by the left singular matrix. constitute, Dimensional scheduling variables , This can be further expanded to: (7) (8) Therefore, the longitudinal LPV model of a supersonic civil aircraft can be converted into a multicellular form as follows: (9) Because a two-dimensional scheduling variable consisting of speed and altitude is selected, This can be further expressed as: (10) in Represents the first vertex in the corresponding vertex system. The weight of each scheduling variable, i.e. (11) For a control system as described in equation (11), the condition for the system to achieve asymptotic stability over a wide range at the equilibrium point under uncontrolled conditions is the existence of a positive definite matrix. This makes the following inequality hold: (12) in, and These are the state transition matrix and control matrix of the vertex system selected by HOSVD. The control law design formula under the PDC framework is expressed in a more concise and clear way as follows: (13) Substituting the above equation into the control system, we obtain a differential equation of the following form: (14) Further rearranging the equations, the controlled system can be written in the following form: (15) in: At this point, the controlled system is equivalent to a free-responding object in an uncontrolled state. The asymptotic stability condition of the system under the PDC control framework can be derived: there exists a positive definite matrix. This makes the following inequality hold: (16) (17) Define a new positive definite matrix and the new controller solution matrix Substituting this into the above equation, we finally obtain the controller design criterion based on the convex hull tensor product model: find a positive definite matrix. and controller solution matrix It satisfies the following matrix inequalities: (18) (19) Given an initial velocity of 170 m / s and an initial altitude of 10,000 meters, within 100 seconds, the velocity uniformly reaches 340 m / s and the altitude reaches 10,600 meters. The control effect is as follows: Figure 4 As shown.

[0113] The modeling method based on linear parameter transformation of multicellular structures employed in this invention systematically reduces and filters the continuous set of operating points in a high-dimensional parameter space by introducing mathematical tools such as singular value decomposition. This method can accurately identify a finite number of vertex systems constituting the entire convex envelope. This transformation brings two advantages: firstly, it greatly simplifies the complex problem of full envelope stability analysis and controller synthesis into a design and verification problem for a few vertex systems, thus significantly reducing the computational burden and design complexity; secondly, and more importantly, it provides a solid theoretical foundation for ensuring the performance of the controller within the entire envelope. Based on convex combination theory, as long as all vertex controllers meet specific performance indicators, it can be ensured that any interior point controller composed of its convex combination meets the same performance throughout the entire flight envelope, thereby fundamentally solving the theoretical deficiencies of traditional interpolation methods.

[0114] The foregoing has provided a detailed description of a supersonic civil aircraft control method, system, and readable storage medium based on multicellular decomposition, as provided in the embodiments of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas; furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

[0115] Certain terms are used in the specification and claims to refer to specific components. Those skilled in the art will understand that hardware manufacturers may use different names to refer to the same component. This specification and claims do not distinguish components based on differences in name, but rather on differences in function. The terms "comprising" and "including" used throughout the specification and claims are open-ended and should be interpreted as "comprising / including but not limited to". "Approximately" means that within an acceptable margin of error, those skilled in the art can solve the technical problem and substantially achieve the technical effect within a certain margin of error. The following descriptions in the specification are preferred embodiments for carrying out this application; however, these descriptions are for the purpose of illustrating the general principles of this application and are not intended to limit the scope of this application. The scope of protection of this application shall be determined by the appended claims.

[0116] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a product or system comprising a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a product or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the product or system that includes said element.

[0117] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0118] The foregoing description illustrates and describes several preferred embodiments of this application. However, as previously stated, it should be understood that this application is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the application concept described herein through the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of this application should be within the protection scope of the appended claims.

Claims

1. A control method for supersonic civil aircraft based on multicellular decomposition, characterized in that, The supersonic civil aircraft control method based on multicellular decomposition includes the following steps: S1: Select the scheduling parameters and their variation range within the wide envelope of the aircraft, perform high-density sampling within the envelope based on the scheduling parameters and their variation range to generate a set of working points, and use a linearization method at each working point to obtain a set of local linear models; S2: Singular value decomposition is applied to the local linear model set to reduce dimensionality, identify the core parameter subspace, and select the polyhedral vertex linear models that constitute the convex envelope. The polyhedral vertex linear models cover the nonlinear dynamics of the entire envelope. S3: For the linear model of each polyhedron vertex, a local state feedback controller is designed by solving linear matrix inequalities using a parallel allocation compensation framework. S4: The local state feedback controller is subjected to convex combination weighted fusion using a continuous weight function to generate a global control law that is applied to the full envelope control of the aircraft.

2. The supersonic civil aircraft control method based on multicellular decomposition according to claim 1, characterized in that, In S1: The scheduling parameters include Mach number, altitude, and angle of attack, and the range of variation covers all operating conditions within the wide envelope of the aircraft. The high-density sampling uses a uniform distribution method to generate dynamic datasets, and the sampling point spacing is set to be below a preset threshold. The linearization method is Jacobian linearization, which obtains a local linear model near the equilibrium point for each operating point; the set of local linear models forms a continuous trajectory in a high-dimensional parameter space, and the approximate accuracy is verified by comparing it with the nonlinear dynamics model of the aircraft.

3. The supersonic civil aircraft control method based on multicellular decomposition according to claim 1, characterized in that, In S2: The singular value decomposition decomposes the dominant dynamic direction of the local linear model set and extracts the subspace corresponding to the dominant singular values. The subspace corresponding to the dominant singular values ​​is the core parameter subspace. The core parameter subspace is a reduced-dimensional space; The minimum vertex set is calculated by applying the convex hull algorithm within the core parameter subspace. The linear model of the polyhedron vertex is composed of the linear model corresponding to the minimum vertex set, and the convex hull precisely encloses the linear model corresponding to all sampled working points.

4. The supersonic civil aircraft control method based on multicellular decomposition according to claim 1, characterized in that, In S3: The parallel allocation compensation framework transforms the global control problem into a set of independent linear matrix inequalities for each vertex. The linear matrix inequalities are constructed based on common Lyapunov functions; The state feedback gain of each vertex is obtained by solving linear matrix inequalities; the local state feedback controller is applied to the corresponding vertex closed-loop system.

5. The supersonic civil aircraft control method based on multicellular decomposition according to claim 1, characterized in that, In S4: The continuous weighting function is a polyhedral membership function, which is associated with the position of the current working point within the convex hull. Calculate the weight values ​​for each vertex for the current working point; The local state feedback controller is weighted and summed according to the weight values.

6. The supersonic civil aircraft control method based on multicellular decomposition according to claim 2, characterized in that, In S1: the dynamic dataset contains the aircraft state and parameters at all sampling working points; the preset threshold is determined based on the nonlinear dynamic characteristics of the aircraft.

7. The supersonic civil aircraft control method based on multicellular decomposition according to claim 3, characterized in that, In S2: the convex hull algorithm traverses the dimensionality-reduced sampling point set; the minimum vertex set is the original working point corresponding to the boundary point of the convex hull.

8. The supersonic civil aircraft control method based on multicellular decomposition according to claim 5, characterized in that, In S4: the weighted summation generates the global control gain corresponding to the current operating point; the global control law is obtained by multiplying the global control gain by the aircraft state.

9. A supersonic civil aircraft control system based on multicellular decomposition, used to implement the supersonic civil aircraft control method based on multicellular decomposition as described in any one of claims 1-8, characterized in that, The supersonic civil aircraft control method based on multicellular decomposition includes the following steps: The point set model acquisition module is used to select the scheduling parameters and their variation range within the wide envelope of the aircraft, perform high-density sampling within the envelope based on the scheduling parameters and their variation range to generate a working point set, and use a linearization method at each working point to obtain a local linear model set; The dimension reduction module is used to apply singular value decomposition to the local linear model set for dimension reduction, identify the core parameter subspace and filter out the polyhedral vertex linear models that constitute the convex envelope, and the polyhedral vertex linear models cover the nonlinear dynamics of the entire envelope. The feedback control module is used to design a local state feedback controller for each polyhedron vertex linear model by using a parallel allocation compensation framework and solving linear matrix inequalities. The weighted fusion module is used to perform convex combination weighted fusion on the local state feedback controller using a continuous weight function to generate a global control law for application to the full envelope control of the aircraft.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the supersonic civil aircraft control method based on multicellular decomposition as described in any one of claims 1-8.

Images