A disturbance-oriented information flow approximate modeling method for flight control system

Abstract

The application discloses a kind of information flow approximate modeling methods of flight control system for disturbance, the field model of flight control system obtained by constructing information flow element model and taking it as modeling paradigm can be expressed independently of underlying implementation logic, so as to get rid of the dependence on specific flight control platform, improve the portability and cross-platform deployment capability of modeling method;By symbolizing the disturbance faced by flight control system, the limitations of existing fault-tolerant control methods relying on type-specific modeling assumptions are effectively broken through, and the universality and generality of disturbance modeling are significantly improved;By establishing chain propagation analysis mechanism of disturbance from the input node of field model to the propagation of control output, the transmission path of disturbance in field model can be tracked and quantitatively analyzed explicitly, and the explainability of abnormal behavior modeling is enhanced.The application is widely used in the field of unmanned aerial vehicle technology.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to an approximate modeling method for information flow in a flight control system oriented towards disturbances. Background Technology

[0002] In recent years, with technological innovation and continuous reduction in manufacturing costs, unmanned aerial vehicles (UAVs) have gradually been widely used in both military and civilian fields. The flight control system, or flight control system for short, is the core component of a UAV. It is responsible for processing data from various sensors, generating and processing telemetry and remote control data, and has a significant impact on the stability of UAV flight, the reliability of data transmission, and the real-time performance of operations. However, flight control systems face various types of disturbances in complex environments. These disturbances include not only environmental interference caused by natural factors (such as wind speed fluctuations and temperature changes), but also component failures within the system (such as sensor failure and actuator degradation), as well as cyberattacks launched by malicious attackers (such as data injection and signal spoofing). These disturbances can directly or indirectly affect the flight stability, navigation accuracy, and mission reliability of the UAV.

[0003] To address these disturbances, some UAV flight control systems employ fault-tolerant control mechanisms. The basic idea is to model specific disturbance types and introduce compensation mechanisms into the control logic. Generally, fault-tolerant controller design can be divided into two categories: passive fault-tolerant control and active fault-tolerant control. Passive fault-tolerant control tolerates faults by abstracting the effects of disturbances as bounded uncertainties without altering the original control structure, thus improving system robustness. In contrast, active fault-tolerant control relies on real-time fault detection and diagnosis to reconfigure the controller in response to disturbances or faults. While these methods can construct effective model representations in specific disturbance scenarios, they still suffer from two key limitations in terms of disturbance characterization and structural abstraction. First, their disturbance modeling relies on idealized assumptions for specific types of disturbances, limiting their generalization to heterogeneous disturbances (especially cyberattacks). Second, their tightly coupled architecture with specific platforms hinders structural generalization and increases the overhead of cross-platform deployment. These limitations highlight the need for a modeling and analysis approach that supports generalized disturbance characterization and structural abstraction independent of platform-specific control architectures. Summary of the Invention

[0004] To address the technical problems of current disturbance-oriented flight control system modeling techniques, such as insufficient generalization of heterogeneous disturbances and high dependence on specific platform control architectures, the present invention aims to provide a disturbance-oriented flight control system information flow approximation modeling method.

[0005] On one hand, embodiments of the present invention include a method for approximate modeling of information flow in a flight control system oriented towards disturbances. This method includes the following steps:

[0006] Using the information flow meta-model as the modeling paradigm, a domain model of the flight control system is established;

[0007] The disturbances faced by the flight control system are symbolically and uniformly represented to obtain heterogeneous disturbance information;

[0008] Embed the heterogeneous perturbation information into the domain model;

[0009] A disturbance propagation analysis is performed on the domain model.

[0010] Furthermore, the information flow meta-model includes unit-type elements, information-type elements, and connection-type elements; the unit-type elements include functional modules, nodes, and transfer functions; the node is the smallest information processing unit executed by the information flow, and the node converts input variables into output variables through the transfer function.

[0011] Furthermore, the establishment of a domain model for the flight control system using the information flow meta-model as the modeling paradigm includes:

[0012] The flight control system is divided into functional modules, and multiple functional modules are identified;

[0013] Identify key code blocks from the functional modules and use these key code blocks as key nodes;

[0014] The key nodes are classified as measurement nodes, execution nodes, control nodes, or aggregation nodes.

[0015] Obtain the relevant variables and transfer functions of the key nodes;

[0016] The domain model is composed of the functional modules, the key nodes, the relevant variables, and the transfer function.

[0017] Furthermore, the flight control system is functionally divided into multiple modules, including:

[0018] Obtain the configuration file of the flight control system;

[0019] Based on the configuration file, the compilation dependencies between the various modules in the flight control system are determined;

[0020] Based on the compilation dependencies, the modules of the flight control system are combined and split to obtain the functional modules.

[0021] Furthermore, identifying key code blocks from the functional modules includes:

[0022] The reverse slicing method is used to trace and analyze the data dependencies of the code in the functional module to identify the key code blocks.

[0023] Furthermore, the method of employing reverse slicing to track and analyze data dependencies within the functional modules includes:

[0024] The output variables in the aforementioned functional module are used as the slice target;

[0025] Starting from the output variable, trace backward along the data dependency path until encountering data from upstream or external instructions;

[0026] Mark the data dependency paths that have been traced in reverse;

[0027] Extract the code block corresponding to the marked data dependency path, and use it as the key code block.

[0028] Furthermore, obtaining the relevant variables and transfer functions of the key node includes:

[0029] Obtain the input and output variables of the key nodes;

[0030] The transfer function is determined based on the input variables and the output variables.

[0031] Furthermore, the step of symbolically and uniformly representing the disturbances faced by the flight control system to obtain heterogeneous disturbance information includes:

[0032] The disturbances faced by the flight control system are represented as input variables or transfer function deviations of the disturbed nodes, and the effects of the disturbances are represented as sign deviation terms in the output variables of the measurement nodes, thus obtaining the heterogeneous disturbance information.

[0033] Furthermore, embedding the heterogeneous perturbation information into the domain model includes:

[0034] The measurement nodes in the domain model are configured based on the sign bias term in the output variables of the heterogeneous perturbation information.

[0035] Furthermore, the perturbation propagation analysis of the domain model includes:

[0036] Gain access to injection attacks;

[0037] The injection attack is modeled as a disturbance to the flight control system and input into the domain model;

[0038] Obtain the chained Jacobian matrix between each of the key nodes in the domain model;

[0039] Based on the chained Jacobian matrix, the propagation process of the disturbance between each of the key nodes is determined.

[0040] The beneficial effects of this invention are as follows: The disturbance-oriented flight control system information flow approximation modeling method in the embodiments, by constructing an information flow meta-model and using it as a modeling paradigm, allows the obtained domain model of the flight control system to be expressed independently of the underlying implementation logic, thereby eliminating dependence on a specific flight control platform and improving the portability and cross-platform deployment capability of the modeling method; by symbolically and uniformly representing the disturbances faced by the flight control system, it effectively overcomes the limitations of existing fault-tolerant control methods that rely on type-specific modeling assumptions, significantly improving the breadth and universality of disturbance modeling; by establishing a chain propagation analysis mechanism for disturbances from the input nodes of the domain model to the control output, it is possible to explicitly track and quantify the transmission path of disturbances within the domain model, enhancing the interpretability modeling of abnormal behavior. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the information flow meta-model in the embodiment;

[0042] Figure 2 This is a schematic diagram of the measurement node, execution node, control node, and aggregation node included in the information flow meta-model in the embodiment;

[0043] Figure 3 This is a schematic diagram illustrating the steps of approximate modeling of the information flow of a flight control system oriented towards disturbances in the embodiment.

[0044] Figure 4 This is a schematic diagram of the framework of the perturbation-oriented flight control system information flow approximation modeling method in the embodiment;

[0045] Figure 5 This is a schematic diagram of the internal information flow architecture of the domain model established based on the PX4 Autopilot flight control system in the embodiment.

[0046] Figure 6 This is a schematic diagram showing the performance comparison between PX4 and the domain model in the nominal scenario during simulation experiments in the embodiment.

[0047] Figure 7 This is a schematic diagram comparing the performance of PX4 and the domain model under wind interference scenarios during simulation experiments in the embodiments.

[0048] Figure 8 This is a schematic diagram comparing the performance of PX4 and the domain model under an actuator failure scenario during simulation experiments in the embodiment.

[0049] Figure 9 This is a schematic diagram comparing the performance of PX4 and the domain model under a network attack scenario during simulation experiments in the embodiment. Detailed Implementation

[0050] I. Shortcomings and Solutions in the Field of Unmanned Aerial Vehicle (UAV) Flight Control System Modeling and Disturbance Characterization

[0051] (1) Insufficient generalization of perturbation representation

[0052] Current fault-tolerant control technologies for UAV flight control systems often rely on highly specific disturbance modeling assumptions, such as treating external wind disturbances as Gaussian noise and modeling sensor failures as fixed biases. This type of specific and static disturbance modeling approach struggles to characterize real-world heterogeneous disturbances, especially highly unstructured or unpredictable cyberattacks. Furthermore, the lack of a unified disturbance modeling framework makes it difficult to systematically analyze the propagation paths of heterogeneous disturbances in control systems, thus limiting their ability to model and evaluate system disturbance response mechanisms.

[0053] (2) Insufficient generalization of control architecture

[0054] Currently, fault-tolerant control methods are typically deeply embedded in the specific platform control architecture, and their control logic is usually highly coupled with the platform implementation, making cross-platform migration difficult. Deploying a fault-tolerant strategy to a new UAV platform often requires refactoring the control law, adjusting the feedback structure, or embedding additional modules. This not only increases the engineering cost of deployment but also affects the practicality and portability of the control system in resource-constrained devices.

[0055] (3) Difficulty in selecting modeling granularity

[0056] Unmanned aerial vehicle (UAV) flight control systems, such as PX4 Autopilot and ArduPilot, typically contain hundreds or even thousands of adjustable parameters and multiple levels of control logic. Their internal information flow and interactions are complex, making comprehensive modeling using traditional methods difficult. Therefore, determining an appropriate granularity of abstraction is crucial. If the granularity is too coarse, key control dependencies may be overlooked, affecting modeling accuracy; conversely, if the granularity is too fine, it may lead to state space explosion, rendering computation infeasible. Therefore, there is an urgent need to construct a flexible and adjustable approximate modeling method that, while preserving the core control semantics of the flight control system, enables controllable abstraction and analysis support for key information flows.

[0057] Information flow analysis offers a novel research approach to address the limitations of existing methods in structural generalization and disturbance representation. Unlike traditional fault-tolerant methods that intervene in low-level control logic, information flow analysis abstracts the system based on how information propagates and transforms among system components. This higher-level modeling perspective avoids dependence on platform-specific control implementations, thereby achieving structural decoupling and enhancing model portability. In addition to supporting structural abstraction, this modeling perspective also supports the generalized representation of heterogeneous disturbances as anomalies in the system's information flow, despite their diverse sources and manifestations. For example, component failures may disrupt information flow through data loss, environmental disturbances may interfere with information flow through random noise, and cyberattacks such as sensor spoofing may manipulate or disrupt information flow. This unified perspective provides an easily analytical framework for representing both traditional disturbances and complex adversarial cyberattacks within a single modeling paradigm.

[0058] II. Information Flow Metamodel

[0059] The information flow meta-model provides a generalizable abstraction method for system information interaction. It offers a unified representation of heterogeneous disturbances and abstracts internal control logic with appropriate granularity. Furthermore, it supports extension to domain-specific models, enabling traceable analysis of the impact of disturbances on control behavior while preserving the control semantics of the flight control system. Therefore, in this embodiment, the information flow meta-model is used as the modeling paradigm for the UAV flight control system.

[0060] The following section introduces relevant knowledge about the information flow meta-model, including element classification, as well as key elements in the information flow meta-model such as nodes, information flow, and transfer functions.

[0061] (I) Element Classification of the Information Flow Meta-Model

[0062] The information flow meta-model includes multiple elements. Based on the characteristics of the control system, the information flow meta-model divides system elements into three categories: unit type, information type, and connection type.

[0063] (1) Unit Class: This class includes elements of various types, such as tasks (e.g., hovering, path planning), entities (e.g., drones, unmanned vehicles), modules (e.g., control modules, dynamics modules), nodes (e.g., measurement nodes, execution nodes), and transfer functions (e.g., P-PID control functions). These elements define the functional and structural units of the system.

[0064] (2) Information elements: These include elements such as commands (e.g., desired position, desired heading), constants (e.g., gravitational acceleration, UAV mass), variables (e.g., measured velocity, measured angle), constraints (e.g., maximum tilt angle, actuator limits), and disturbances (e.g., component failure, network attack). These elements constitute the core data for transmitting control and status information between nodes.

[0065] (3) Connection class: Defines the interaction mechanism within the system, including the information flow, association, communication and control relationships between unit class elements.

[0066] Based on the classification, the membership relationships between elements are established, and finally, a reference meta-model for information flow analysis, namely the information flow meta-model, is constructed, such as... Figure 1 As shown.

[0067] (II) Key Elements in the Information Flow Meta-Model

[0068] (1) Node classification

[0069] In the information flow meta-model, a node is defined as the smallest unit of information processing, which transforms input variables into output variables through transfer functions. These nodes constitute the basic unit blocks for information flow execution, and the interconnection and combination of node-level transformations collectively define the system's control logic. For example... Figure 2 As shown, based on their role in the control system, the nodes in the information flow meta-model can be divided into four functional types: measurement nodes, execution nodes, control nodes, and aggregation nodes.

[0070] Among them, measurement nodes such as Figure 2 As shown by the orange square in the upper left corner, it can sense physical quantities and convert them into digital information, serving as the source of the information flow. Execution nodes are as follows: Figure 2 As shown by the green square in the upper right corner, it can receive control commands and issue instructions to change the operating state of the actuator. The control node is as follows: Figure 2 As shown by the blue square in the lower left corner, it can convert input into control output through a predefined control algorithm. Aggregation nodes are as follows... Figure 2 As shown by the red square in the lower right corner, it can integrate multiple input streams or distribute data to multiple outputs, thereby achieving information fusion and parallel propagation within the system.

[0071] (2) Information flow

[0072] Reference Figure 2 The dependencies between nodes can be represented by the direction of information flow and modeled using an association matrix:

[0073]

[0074] Where T is the correlation matrix, N m Let t be the number of nodes. If the input of node j depends on the output of node k, then t jk =1; otherwise t jk =0.

[0075] In flight control systems, control signals are based on estimated measured states x. mGenerated. Suppose there are K control nodes in total, and the control signal generated by the k-th node is denoted as u. k Then the control output vector u can be defined as:

[0076]

[0077] If the structure and function of the control system are known, then estimate the measured value x. m The information flow to the control output u can be modeled as the following nonlinear mapping relationship:

[0078] x m (N)→u(N)=Φ(x m (N)) (3)

[0079] Wherein, Φ represents the internal information conversion process of the system determined by the control logic.

[0080] (3) Node transfer function

[0081] The transfer function is the fundamental representation of the dynamic characteristics of a node, used to describe how each node processes input variables into corresponding outputs. For the i-th node, let its input variable be... The output variable is The transfer function of this node is defined as follows:

[0082]

[0083] However, when a node is disturbed, its input or transfer function may change, and the disturbed output can be expressed as:

[0084]

[0085] in, δF typically represents the input deviation caused by environmental disturbances. i This typically represents a change in the transfer function caused by component failure. A network attack, however, can interfere with both the input and transfer functions, individually or simultaneously. For example, data injection attacks often cause… Code injection attacks, on the other hand, generate δF. i Therefore, considering that cyberattacks can introduce arbitrary forms of perturbation and have uncertain, discontinuous, or nonlinear characteristics, this approach does not... and δF i Apply any structural assumptions.

[0086] III. Approximate Modeling Method for Information Flow in Flight Control Systems Oriented to Disturbances

[0087] To address the shortcomings of traditional fault-tolerant control methods in terms of disturbance representation and structural abstraction, this embodiment provides an approximate modeling method for information flow in flight control systems, based on an information flow meta-model. This method abstracts the control logic of the flight control system into a composable information flow, achieving structural generalization and supporting a unified representation of heterogeneous disturbances. Based on the domain model instantiated by this method, the propagation path of disturbances in the control logic and their impact on system behavior can be traced and analyzed, thereby approximating the dynamic response of a real flight system under disturbance scenarios.

[0088] Reference Figure 3 The approximate modeling method for information flow in a flight control system oriented towards disturbances includes the following steps:

[0089] S1. Using the information flow meta-model as the modeling paradigm, establish the domain model of the flight control system;

[0090] S2. Symbolically represent the disturbances faced by the flight control system to obtain heterogeneous disturbance information;

[0091] S3. Embed heterogeneous perturbation information into the domain model;

[0092] S4. Perform perturbation propagation analysis on the domain model.

[0093] In this embodiment, Figure 3 The framework of the disturbance-oriented flight control system information flow approximation modeling method is as follows: Figure 4 As shown, the framework mainly consists of three parts: First, in step S1, the information flow meta-model (referred to as the meta-model) is used as the abstract modeling paradigm to define three key elements, including unit classes, information classes, and connection classes; then, based on the modeling paradigm of the meta-model, a specific domain model of the flight control system is constructed through steps such as functional module division, key node identification, and variable and transfer function mapping; next, step S2 is executed, and according to the definition of the information flow meta-model, the disturbances faced by the flight control system (specifically including network attacks, component failures, and environmental interference) are uniformly represented as measurable information flow deviations of the system, that is, heterogeneous disturbance information is obtained; finally, steps S3-S4 are executed to analyze how the disturbances propagate along the information flow and change the control output in the domain model.

[0094] The information flow domain model of the flight control system to be established in step S1 (hereinafter referred to as the domain model) is equivalent to a concrete instantiation of the information flow meta-model. The domain model extends the meta-model, enabling it to perform fine-grained representation of control logic and information dependencies in the actual platform.

[0095] Since the system elements in the metamodel are divided into three categories: unit classes, information classes, and connection classes, this invention focuses on key elements that accurately reflect the control logic and information dependencies when constructing a specific domain model for an unmanned aerial vehicle (UAV) flight control system. These key elements include modules, nodes, and transfer functions in the unit classes, and variables in the information classes. Therefore, the construction of the domain model in step S1 mainly includes three stages: functional module partitioning, key node identification, and variable-transfer function mapping.

[0096] Based on the above principles, in this embodiment, when performing step S1, which is to establish the domain model of the flight control system using the information flow meta-model as the modeling paradigm, the following steps can be specifically performed:

[0097] S101. Divide the flight control system into functional modules and determine multiple functional modules;

[0098] S102. Identify key code blocks from functional modules and use these key code blocks as key nodes;

[0099] S103. Obtain the relevant variables and transfer functions of key nodes;

[0100] S104. Based on functional modules, key nodes, relevant variables, and transfer functions, construct a domain model.

[0101] Among them, step S101 belongs to the functional module division stage, step S102 belongs to the key node identification stage, and step S103 belongs to the variable-transfer function mapping stage.

[0102] (1) Step S101: Functional module division

[0103] In a UAV flight control system, each module is responsible for executing a specific control function, and the information flow between modules determines the system's control behavior. Therefore, effective module division forms the basis for domain model construction. Reasonable division can reduce unnecessary inter-module dependencies and improve information traceability.

[0104] To reveal the information flow relationships between modules, this embodiment analyzes the system build configuration file CMakeLists.txt and extracts the compilation dependencies of custom modules in the system architecture. Modules that frequently exchange data or are tightly coupled in computation are grouped into the same functional module, while functionally independent parts with limited interaction are split into different modules. For example, in the default PX4 architecture, mc_pos_control and mc_att_control are implemented as two independent modules, but there is a tight information coupling between them, that is, the expected acceleration output by the former serves as the input of the latter. Therefore, to reduce modeling complexity, they can be treated as a single module in the domain model.

[0105] (2) Step S102: Key Node Identification

[0106] After completing the functional module division, the next step is to identify key nodes, i.e., code blocks that significantly impact system stability, task execution, or information integrity during system operation. To identify key code blocks in the system, this embodiment employs the reverse slicing method from static program analysis, analyzing data dependencies through tracing. The core idea of ​​the reverse slicing method is to start with the output variable (e.g., actuator_controls) of a node (e.g., an actuator) and trace backwards along the data dependency path to identify its upstream computational logic, terminating when data or external command input from the upstream of this node (e.g., the upstream of the actuator—a sensor) is encountered. All data dependency paths related to the target variable are marked, and the corresponding code blocks are extracted for analysis. Based on their functional roles defined in the meta-model paradigm, these code blocks are classified as aggregation nodes, control nodes, measurement nodes, or execution nodes.

[0107] In this embodiment, the key node identification performed in step S102 is explained using the algorithm contained in the flight control system as an example.

[0108] Algorithm 1 implements a reverse slicing process based on the target variable tv and the flight control source code SC to identify code blocks that affect the calculation of the target variable. These code blocks are key code blocks, referred to as key nodes in the domain model.

[0109] For Algorithm 1, the variable analysis list W, the set of visited variables V, and the output set RC are first initialized (lines 2–4). Then, variable v is sequentially retrieved from W; if v has not yet been visited, it is marked and further analyzed (lines 5–9). The assignment statement for variable v is found using the FindAssignment(v,SC) function (line 10), and the source variable and its corresponding code block are extracted (lines 11–12). If the source variable comes from a sensor subject or navigator, the tracking is terminated; otherwise, the source variable is added back to W for further analysis (lines 13–15). Finally, the algorithm returns the set RC containing all key nodes (line 22).

[0110]

[0111]

[0112] (3) Step S103: Variable-transfer function mapping

[0113] During the reverse slicing process, the input and output variables of key nodes and their corresponding transfer functions are synchronously mapped to the domain model to define their information dependencies. As shown in Algorithm 1, starting from line 16, each identified key node (i.e., key code block) is traversed. The assignment statements within each node are analyzed, the output variables on the left and the input variables on the right are extracted, and stored in the variable set Var (lines 17–19). Subsequently, the internal computational logic of each node is parsed, and the corresponding transfer function is derived based on the input-output relationship and stored in the transfer function set TF (lines 20–21). Finally, the variable set Var and the transfer function set TF are returned together as the algorithm output (line 22).

[0114] Taking the control node in the PX4 system (a specific flight control system) as an example, the input variables of the yaw control node include att_sp.yaw_body, yaw_sp_move_rate, yaw_body, and yaw_move_rate, and the output variable is actuator_controls[2]. The transfer function corresponding to this node can be extracted as a cascaded P-PID equation. In addition to the input and output variables, each key node may also be associated with other types of information elements, such as constants, constraints, and commands. These elements are also integrated into the domain model to improve its information dependency representation.

[0115] (4) Specific examples of the domain model established in steps S101-S104

[0116] In this embodiment, based on the specific flight control system PX4 Autopilot, a concrete instantiation of the domain model established by performing steps S101-S104 of the flight control system is demonstrated. Specifically, according to the classification principles of meta-models and the domain model construction methods of the aforementioned steps S101-S104, its functional modules, key nodes, related variables, and transfer functions are described in detail.

[0117] (4.1) Functional modules and key nodes

[0118] Figure 5 The internal information flow architecture of the instantiated domain model is illustrated. This model primarily consists of four modules: a navigation module, a control module, a dynamics module, and a measurement module. These modules are connected in a closed-loop structure: First, the navigation module converts task commands into control setpoints; then, the control module calculates actuator commands based on the setpoints and sensor feedback; the dynamics module receives these commands and executes them via motors; finally, the measurement module collects sensor data, providing feedback for closed-loop control. This sequence of information flow constitutes the basic operational logic of the system.

[0119] Based on the reverse slicing method used in step S102, this embodiment selects the system output variable actuator_controls as the slicing target.

[0120] Figure 5 The identified key nodes are displayed and categorized into four classes based on the meta-model. Aggregation nodes include the navigator, MAVLink receiver, and Extended Kalman Filter (EKF), used to integrate task inputs, communication data, and fuse multi-sensor measurements to generate state estimates, respectively. Control nodes encompass six-DOF P-PID control nodes and control assignment nodes. Actuation nodes include motor actuation nodes and six motion nodes related to translational and rotational dynamics. Measurement nodes include gyroscopes, accelerometers, magnetometers, GPS, and barometers, providing raw inertial and environmental data for state estimation.

[0121] (4.2) Node variables and transfer functions

[0122] The domain model is ultimately formed by specifying the transformation behavior (i.e., transfer functions) of key nodes. These functions describe how input variables are processed to generate output variables, reflecting the basic control logic of the system. This embodiment focuses on analyzing four representative types of nodes: P-PID control nodes, control assignment nodes, motor execution nodes, and motion nodes.

[0123] Taking height control as an example, the P-PID control node adopts a cascaded control structure: the outer loop input is the desired height z. d (t) and the measured height z(t), the inner loop input is the desired vertical velocity v zd (t) and the measured vertical velocity v z (t), the output variable is the total thrust command u1(t). The outer loop uses a proportional control function to convert the height error into the desired vertical velocity:

[0124] v zd (t)=K P,z ·e z (t)=K P,z ·(z d (t)-z(t)) (6)

[0125] Among them, K P,z For the proportional gain, the calculated v zd (t) is used as the setpoint for the inner loop speed control. The inner loop uses a PID control function to generate thrust commands:

[0126]

[0127] in, PID gain K PK I K D These determine the proportional, integral, and derivative response characteristics, respectively. The control distribution node maps the overall control command to individual actuator instructions. The input variables are four control quantities: total thrust u1 and torque commands in three directions u2 (roll), u3 (pitch), and u4 (yaw). The output variables are the squares of the speeds of the four motors.

[0128] For a symmetrical cruciform quadcopter, the distribution relationship is as follows:

[0129]

[0130] Where K T K is the thrust coefficient. d Here, l represents the drag coefficient, and λ represents the arm length. This conversion ensures that control commands are appropriately distributed to the motors according to the aircraft's geometry.

[0131] The motor actuation nodes and motion nodes in the dynamics module describe how the actuator generates the actual motion. The input variable for each motor node is the square of the rotational speed. The output variable is the corresponding thrust f i It is given by the following formula:

[0132]

[0133] The total thrust drives the aircraft to produce translational motion (lateral, longitudinal, and vertical). The translational motion nodes receive the thrust f. i Current attitude angles φ, θ, ψ, and linear velocity v x v y v z As input, the output variable is linear acceleration. The calculation formula is as follows:

[0134]

[0135] Where m is the mass of the aircraft and g is the gravitational constant. The above function describes how thrust is converted into inertial acceleration, taking into account attitude and aerodynamic drag.

[0136] The rotational motion node describes the dynamic behavior in the roll, pitch, and yaw directions, with the input being the thrust f. i Given angular velocities p, q, and r, the output is angular acceleration. The calculation is performed using Euler's dynamics formula:

[0137]

[0138] Among them, I xx I yy I zz, , represent the moment of inertia of the aircraft in three directions, and s is the distance from the motor to the aircraft's center of mass. These functions describe how the aircraft generates angular acceleration based on the thrust difference between the motors.

[0139] The domain model constructed using the aforementioned transfer function clearly characterizes the entire process from control commands to actuator signals, and then to changes in physical state. This formalized structure provides a structural foundation for subsequent analysis of disturbance propagation.

[0140] To enhance the modeling and generalization capabilities of flight control systems when facing heterogeneous disturbances, this embodiment proposes a symbolic disturbance modeling and analysis method. Specifically, in step S2, which involves symbolically and uniformly representing the disturbances faced by the flight control system to obtain heterogeneous disturbance information, the heterogeneous disturbances are first uniformly represented. Then, step S3 is executed to embed the heterogeneous disturbances into the flight control domain model and analyze the propagation path and impact mechanism of the disturbances in the control information flow.

[0141] (5) Steps S2-S3: Unified representation of heterogeneous perturbations and embedding them into the domain model

[0142] In the meta-model, each key information processing unit within the control system is abstracted as a node, whose definition includes input variables. Transfer function F i And by formula The calculated output variable is shown in formula (4). Assuming the system consists of L serial nodes, then in an ideal scenario, the measured output x m (t) is derived from the expected value x d (t) and the output x from the previous time step m (t-1) The combined result after transforming all nodes in sequence:

[0143]

[0144] According to formula (5), heterogeneous disturbances in the system may affect input variables or internal transfer functions. When network attacks, component failures, and environmental interference coexist, node N... i The disturbed output can be expressed as:

[0145]

[0146] To analyze the resulting bias, a first-order Taylor expansion of the formula is performed:

[0147]

[0148] in, Let represent the local Jacobian matrix. The above equation quantitatively describes the impact of input perturbations and transfer function changes on the output of a single node. This impact then propagates downwards along the information flow. According to the chain rule, from node N... i Propagate down to measurement node N L During the process, the deviation caused by the disturbance at the measurement output is:

[0149]

[0150] Substituting formula (14) into formula (15), we get:

[0151]

[0152] in, Indicates input disturbance For the final output x m Cumulative sensitivity, This describes the change of the function δF i The impact of its spread.

[0153] When multiple nodes When simultaneously affected by disturbances, the total deviation ε of the measurement output can be expressed by the superposition effect of all disturbed nodes as follows:

[0154]

[0155] in, This represents the higher-order residuals that are ignored in the first-order approximation.

[0156] Formula (17) provides a unified characterization of heterogeneous disturbances. Its core idea is to represent the impact of the disturbances as a sign bias term ε in the measurement output. This bias term serves as a standard form for embedding heterogeneous disturbances into a specific information flow domain model, enabling these disturbances to be formally modeled and computed within the model.

[0157] Therefore, step S2 involves representing the disturbance faced by the flight control system as the input variable or transfer function deviation of the disturbed node, and representing the impact of the disturbance as the sign deviation term in the output variable of the measurement node to obtain heterogeneous disturbance information. In step S3, the sign deviation term is used to configure the output variable of the measurement node in the domain model, thereby embedding the heterogeneous disturbance information into the domain model.

[0158] (6) Step S4: Perturbation Propagation Analysis in Domain Model

[0159] The unified disturbance representation method proposed in step S2 enables any disturbance source to be symbolically mapped as a measurement deviation term, thereby achieving closed-loop traceability from the disturbance source to the control input. Compared to traditional fault-tolerant control methods that mainly focus on component failures and environmental disturbances, this method further incorporates network attacks into a unified and computable modeling structure, significantly expanding its applicability.

[0160] To verify the computational feasibility and scalability of this characterization method, this embodiment uses a network attack scenario as a representative example of the disturbances faced by a UAV flight control system. It illustrates how to instantiate an accelerometer injection attack in a specific flight control system like the PX4 system within the domain model and approximate its impact on downstream control output. In this scenario, the attacker injects forged data into the accelerometer measurement stream. This attack is modeled as an input bias at the accelerometer measurement node N1, mathematically expressed as:

[0161]

[0162] Here, δa(t) represents the altered acceleration vector. This perturbation is then propagated to EKF node N2, whose output is the estimated state vector. Includes position, velocity, attitude angle, and angular velocity:

[0163]

[0164] According to formula (19), the propagation process of this disturbance can be expressed as:

[0165]

[0166] Among them, J 1:2 Let δF1 represent the chained Jacobian matrix from the accelerometer node N1 to the EKF node N2. Since the transfer functions of the nodes remain unchanged, δF1 = 0. This chained Jacobian can be further decomposed into:

[0167]

[0168] Here, F1 represents the accelerometer's transfer function, which can usually be approximated as an identity mapping, and its derivative is the identity matrix. F2 is the transfer function of the EKF node, and its partial derivative with respect to the acceleration input yields the Jacobian matrix. Therefore, the final perturbation term can be expressed as:

[0169] ε(t)=J EKF ·δa(t) (22)

[0170] The following uses the height control loop of PX4 as an example to evaluate the impact of this disturbance on the control output. This control loop employs a cascaded structure of an outer-loop position controller and an inner-loop speed controller. The controllers rely on z(t) and v in the state estimate. z (t), respectively corresponding to x m The 3rd and 6th elements in (t). After being disturbed, the state estimate becomes:

[0171]

[0172] The outer loop controller is used to calculate the desired vertical velocity, and its transfer function (see Equation (6)) is updated as follows:

[0173]

[0174] The inner loop controller is used to calculate the thrust command, and its transfer function (see Equation (7)) is updated as follows:

[0175]

[0176] Therefore, the deviation of the thrust command output due to the disturbance is:

[0177]

[0178] Substituting the symbolic representation of the disturbance term (see formula (22)), we get:

[0179]

[0180] The final control deviation term is expressed as:

[0181]

[0182] in, and J EKF Lines 3 and 6.

[0183] The above derivation formally describes the propagation process of sensor injection attacks along the information flow path and derives an analytical expression of its impact on control output. This demonstrates that the domain model obtained through steps S1-S3 is not only applicable to environmental disturbances and component failures, but also capable of symbolic modeling and traceable analysis of potential disturbances to UAV flight control systems, such as cyberattacks, thereby effectively achieving a more generalizable disturbance representation.

[0184] IV. Simulation Experiment

[0185] In this embodiment, a series of simulation experiments were designed and carried out to compare the mainstream open-source flight control system PX4Autopilot with the flight control system information flow domain model to verify the effectiveness of the modeling method proposed in this invention.

[0186] (1) Experimental setup

[0187] The flight control system used in this experiment consists of PX4 Autopilot, Gazebo simulator, and Robot Operating System (ROS), running on a virtual machine configured with 64-bit Ubuntu, equipped with a quad-core 2.6GHz processor and 8GB of memory. The UAV model used in the PX4 simulation is an Iris quadcopter, and its detailed parameters are shown in Table 1.

[0188] Table 1. Parameters of the Iris Quadrotor

[0189]

[0190]

[0191] A cross-process static backward slice analysis was performed on the intermediate representation of LLVM 13.0 using the SVF 3.0 tool to construct the domain model. The instantiated domain model was implemented based on Python 3.9, and its configured PID gain parameters are listed in Table 2. All simulation experiments were performed on a desktop computer equipped with an Intel Core i7-10700 (2.90GHz) processor and 16GB of memory.

[0192] Table 2 PID Gain Parameters for the Domain Model

[0193]

[0194] (2) Effect verification

[0195] In this embodiment, four representative typical scenarios are selected to systematically evaluate the effectiveness and generalization ability of the proposed domain model. In all experiments, the UAV performs the same task: take off from the initial position (0,0,0)m, fly to the target position (3,3,3)m, and achieve stable hovering at that point.

[0196] To achieve comparability assessments across different systems, PX4 and the domain model each introduce disturbances using their own adaptive methods. In PX4, actuator failures, wind interference, and sensor attacks are directly implemented by modifying the simulation environment or dynamic model. In the domain model, disturbances are modeled as symbolic biases within the measurement information and injected into the input variables of the measurement nodes. Details of the disturbance modeling methods and propagation mechanisms are provided in the embodiments of steps S2-S4. This symbolic modeling method offers good computational feasibility and propagation traceability, supporting approximate modeling of the impact of real disturbances.

[0197] The evaluation covers four scenarios: a disturbance-free nominal scenario, a wind-interference scenario, an actuator failure scenario, and a sensor attack scenario. The control performance of PX4 and the domain model is compared and evaluated using the following three metrics: overshoot, which refers to the maximum deviation of the system response from the set value, used to measure the aggressiveness of the response; settling time, which refers to the time required for the system output to enter and remain within ±2% of the set value; and root mean square error, which refers to the average deviation between the actual trajectory and the expected trajectory of the system, used to reflect the overall tracking accuracy.

[0198] (1) Nominal scenario without disturbance

[0199] In a perturbation-free nominal scenario, this embodiment performs step response analysis on a system based on PX4 and a domain model to verify the domain model's ability to approximate real flight control behavior. Figure 6 The response curves of both are shown on the six control channels, and the corresponding quantitative performance indicators are shown in Table 3.

[0200] from Figure 6 It can be seen that the response trends of the domain model in each axis are basically consistent with those of PX4. In terms of position control, the domain model and PX4 exhibit similar upward trends and steady-state values ​​in the X, Y, and Z axes. Although there is some overshoot and response delay in the X and Y directions, their root mean square errors are almost equivalent to those of PX4. In terms of attitude control, the domain model can accurately reproduce the directional stability and dynamic convergence characteristics of roll and pitch of PX4, with negligible differences in root mean square errors. In summary, although the domain model has slight deviations in some dynamic indicators, its overall control performance remains highly consistent with PX4, verifying its good structural inheritance and dynamic response consistency with the flight control system under nominal conditions.

[0201] Table 3 Comparison of Control Indicators under Nominal Scenarios

[0202]

[0203] (2) Wind interference scenario

[0204] This scenario simulates the impact of environmental wind disturbances on the system's control performance. Specifically, a constant horizontal gust along the positive Y-axis is applied at t=10s and persists throughout the subsequent flight to simulate wind disturbances that may occur in reality. Figure 7 The response curves of PX4 and the domain model under wind disturbance conditions are shown, and Table 4 summarizes the corresponding performance indicators.

[0205] As shown in Table 4, the domain model's response performance is similar to that of PX4 across all axes, especially in Z-axis height control and yaw angle adjustment, where there is almost no difference between the two, with both root mean square error (RMSE) and overshoot approaching zero. On the Y-axis, the principal axis of wind disturbance, although the domain model is slightly higher than PX4 in overshoot and RMSE, its overall trend and control stability are comparable to PX4. On the X-axis, the non-primary lateral disturbance direction, the control performance of both models is almost identical. These results verify that the constructed domain model can approximate the control response of PX4 under wind disturbance conditions.

[0206] Table 4 Comparison of Control Indicators under Wind Interference Scenarios

[0207]

[0208] (3) Actuator failure scenarios

[0209] To evaluate the ability of the domain model to approximate the PX4 control response under actuator failure scenarios, this experiment injected a mild fault into the thrust command for 30 seconds during hovering. Figure 8 The response curves of each axis of PX4 and the domain model under this fault condition are shown, and Table 5 lists the corresponding control performance indicators.

[0210] from Figure 8 As can be seen, although the fault caused a certain deviation in the system trajectory, the response trend of the domain model in all axes is similar to that of PX4, and the dynamic process changes synchronously. In the X and Y axes, the overshoot and root mean square error of the domain model and PX4 are close, and their settling times are similar, exhibiting consistent closed-loop control characteristics. In the Z axis, due to the direct impact of the fault on the thrust channel, the domain model's response shows a more obvious dynamic deviation, but overall it can still approximate the system behavior of PX4 and maintain stable control. In the yaw angle channel, the difference in control accuracy between the two is minimal. Overall, the domain model can effectively reproduce the impact of actuator faults on the dynamic performance of the control system and has good fault modeling capabilities.

[0211] Table 5 Comparison of control indicators under actuator failure scenarios

[0212]

[0213]

[0214] (4) Sensor attack scenarios

[0215] To verify the ability of the domain model to approximate real control behavior when subjected to sensor attacks, this experiment injected an attack signal lasting 30 seconds into the accelerometer observations during the steady-state hovering phase. Figure 9 The response variations of PX4 and the domain model across six axes are shown, and Table 6 summarizes the corresponding control performance metrics.

[0216] The experimental results show that, under the influence of the attack, both systems maintain stable control along all axes, and their trajectory responses exhibit a consistent trend. Specifically, on the X and Y translation axes, the differences between the neighborhood model and PX4 in terms of overshoot and root mean square error are small, and their response characteristics are highly similar. In Z-axis and yaw control, although the neighborhood model shows slightly amplified response fluctuations, it still maintains good dynamic consistency with PX4 overall. These results demonstrate that the neighborhood model can effectively simulate the disturbance effect of sensor attacks on observation information and achieve dynamic approximation of PX4 control behavior, verifying the approximate modeling capability of the symbolic disturbance modeling and propagation method proposed in this study.

[0217] Table 6 Comparison of Control Indicators under Network Attack Scenarios

[0218]

[0219] (5) Efficiency verification

[0220] To evaluate the performance overhead of the proposed model, this embodiment compares two implementation schemes: PX4 and the domain model of this invention. The evaluation was conducted in a unified simulation environment, using three efficiency metrics: lines of code, average execution time, and CPU utilization. Table 7 summarizes the corresponding performance results.

[0221] Given that PX4 comprises multiple components including main control logic, simulation environment, task scheduling framework, and peripheral drivers, to ensure a fair comparison, we only counted the number of lines of code in its main control logic corresponding to the domain model in its src directory, which is approximately 404,000 lines. In contrast, the domain model contains only 2,797 lines of code, with a concise structure and simple implementation. In terms of runtime performance, the domain model's average execution time per control cycle is 0.6 milliseconds, better than PX4's 0.7 milliseconds. PX4's average CPU utilization is 46.5%, while the domain model's is only 36.38%. Considering all three evaluation metrics, the domain model is more lightweight in terms of resource consumption and is suitable for deployment on resource-constrained platforms.

[0222] Table 7 Comparison of Experimental Costs

[0223]

[0224] V. Conclusion

[0225] The disturbance-oriented flight control system information flow approximation modeling method in this embodiment has the following characteristics:

[0226] (1) The information flow meta-model paradigm used in step S1 for modeling flight control system abstracts flight control logic into an information flow network, defines a system structure consisting of unit classes (such as modules, nodes), information classes (such as instructions, variables, disturbances) and connection classes (such as information flow, associations), and supports composable modeling of control logic.

[0227] (2) Step S1 implements the method for constructing the information flow domain model of the flight control system: Based on the information flow meta-model, the domain model is constructed through steps such as functional module division, key node identification, and variable-transfer function mapping. Key node identification adopts the reverse slicing algorithm to trace data dependencies along the control path, extract key variables related to control and their corresponding transfer functions, and use them to construct a structured information flow representation.

[0228] (3) Step S2 realizes a unified representation method for heterogeneous disturbances: by defining the input variable bias (δξ) and the transfer function bias (δF), component failures, environmental interference and network attacks are uniformly abstracted into node-level information flow offset terms, thereby realizing formal modeling and representation of heterogeneous disturbances.

[0229] (4) Step S4 implements the symbolic disturbance propagation analysis method: using the Jacobi chain rule, starting from the disturbance source node, the disturbance propagation process is derived along the system information flow dependency path, and the impact of the disturbance on the measurement results and control output is formally represented, realizing the traceability modeling and quantitative analysis of disturbance propagation in flight control.

[0230] (5) By executing steps S1-S4, a lightweight flight control system information flow domain model is realized. It has the characteristics of simple structure and controllable modeling granularity. It can realize the rapid evaluation of system response under disturbance conditions without relying on real flight control code. It is suitable for the safety verification, adversarial evaluation and simulation optimization tasks before deployment of control systems.

[0231] The disturbance-oriented flight control system information flow approximation modeling method in this embodiment has the following technical effects:

[0232] (1) Enhance the generalization ability of disturbance representation: By abstracting environmental interference, component failure and network attack into symbolic bias in information flow, the limitations of existing fault-tolerant control methods that rely on type-specific modeling assumptions are effectively overcome, and the breadth and universality of disturbance modeling are significantly improved.

[0233] (2) Platform-oriented decoupling structural abstraction: By constructing an information flow meta-model and using it as a modeling paradigm, the system control structure can be expressed independently of the underlying implementation logic, thereby getting rid of the dependence on specific flight control platforms (such as PX4, ArduPilot) and improving the portability and cross-platform deployment capability of the modeling method.

[0234] (3) Traceable modeling of disturbance propagation: By establishing a chain propagation analysis mechanism for disturbances from input nodes to control outputs, the transmission path of disturbances in the system can be explicitly tracked and quantitatively analyzed, which enhances the interpretability modeling of abnormal behavior.

[0235] (4) Lightweight implementation with controllable granularity: By abstracting the control logic and information dependencies at an appropriate granularity, the problem of state space dimension explosion caused by traditional fine modeling is avoided. While retaining the original control logic, the computational complexity is effectively reduced, making the model lightweight and suitable for resource-constrained embedded systems.

[0236] (5) Engineering practicality of approximate modeling: Under multiple typical disturbance scenarios (such as wind disturbance, fault, attack), the domain model shows a high degree of consistency with the real flight control system (such as PX4) in terms of dynamic response and tracking performance, which verifies that the proposed method has good engineering application value while maintaining modeling accuracy.

[0237] The disturbance-oriented flight control system information flow approximation modeling method in this embodiment can be executed by a computer. A computer capable of executing the disturbance-oriented flight control system information flow approximation modeling method possesses all the technical effects of the disturbance-oriented flight control system information flow approximation modeling method.

[0238] It should be noted that, unless otherwise specified, when a feature is referred to as "fixed" or "connected" to another feature, it can be directly fixed or connected to the other feature, or indirectly fixed or connected to the other feature. Furthermore, the descriptions of "upper," "lower," "left," and "right" used in this disclosure are only relative to the relative positional relationships of the components of this disclosure in the accompanying drawings. The singular forms "a," "an," and "the" used in this disclosure are also intended to include the plural forms, unless the context clearly indicates otherwise. Moreover, unless otherwise defined, all technical and scientific terms used in this embodiment have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this embodiment specification is only for describing particular embodiments and is not intended to limit the invention. The term "and / or" as used in this embodiment includes any combination of one or more of the associated listed items.

[0239] It should be understood that although the terms first, second, third, etc., may be used to describe various elements in this disclosure, these elements should not be limited to these terms. These terms are only used to distinguish elements of the same type from each other. For example, a first element may also be referred to as a second element without departing from the scope of this disclosure, and similarly, a second element may also be referred to as a first element. The use of any and all instances or exemplary language (“e.g.,” “such as,” etc.) provided in this embodiment is intended only to better illustrate embodiments of the invention and, unless otherwise required, does not impose a limitation on the scope of the invention.

[0240] It should be recognized that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium. The method can be implemented using standard programming techniques—including a non-transitory computer-readable storage medium configured with a computer program, wherein such a storage medium causes the computer to operate in a specific and predefined manner—according to the methods and drawings described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit (ASIC).

[0241] Furthermore, the procedures described in this embodiment can be performed in any suitable order unless otherwise indicated by this embodiment or clearly contradicted by the context. The procedures (or variations and / or combinations thereof) described in this embodiment can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. A computer program includes multiple instructions executable by one or more processors.

[0242] Furthermore, the method can be implemented in any suitable type of computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices, etc. Aspects of the invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein. Furthermore, the machine-readable code, or portions thereof, can be transmitted via wired or wireless networks. The invention of this embodiment includes these and other different types of non-transitory computer-readable storage media when such media comprises instructions or programs that implement the steps above in conjunction with a microprocessor or other data processor. When programmed according to the methods and techniques of the invention, the invention also includes the computer itself.

[0243] A computer program can be applied to input data to perform the functions of this embodiment, thereby transforming the input data to generate output data stored in non-volatile memory. The output information can also be applied to one or more output devices, such as a display. In a preferred embodiment of the invention, the transformed data represents physical and tangible objects, including specific visual depictions of physical and tangible objects generated on the display.

[0244] The above are merely preferred embodiments of the present invention. The present invention is not limited to the above-described embodiments. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention, as long as they achieve the technical effects of the present invention by the same means, should be included within the scope of protection of the present invention. Within the scope of protection of the present invention, the technical solutions and / or implementation methods can have various modifications and variations.

Claims

1. A method for approximate modeling of information flow in a flight control system oriented towards disturbances, characterized in that, The disturbance-oriented flight control system information flow approximation modeling method includes: Using the information flow meta-model as the modeling paradigm, a domain model of the flight control system is established; The disturbances faced by the flight control system are symbolically and uniformly represented to obtain heterogeneous disturbance information, including: The disturbances faced by the flight control system are represented as input variables or transfer function deviations of the disturbed nodes, and the effects of the disturbances are represented as sign deviation terms in the output variables of the measurement nodes, thereby obtaining the heterogeneous disturbance information. The heterogeneous perturbation information is embedded into the domain model, and the measurement nodes in the domain model are configured according to the sign bias term in the output variables of the heterogeneous perturbation information. Perform perturbation propagation analysis on the domain model; The establishment of a domain model for the flight control system using the information flow meta-model as the modeling paradigm includes: The flight control system is divided into functional modules, and multiple functional modules are identified; Key code blocks were identified from the functional modules, including: The output variables in the aforementioned functional module are used as the slice target; Starting from the output variable, trace backward along the data dependency path until encountering data from upstream or external instructions; Mark the data dependency paths that have been traced in reverse; Extract the code block corresponding to the marked data dependency path as the key code block; The key code block is used as the key node; The key nodes are classified as measurement nodes, execution nodes, control nodes, or aggregation nodes. Obtain the relevant variables and transfer functions of the key nodes; The domain model is composed based on the functional modules, the key nodes, the relevant variables, and the transfer function. The perturbation propagation analysis of the domain model includes: Acquisition injection attack; The injection attack is modeled as a disturbance to the flight control system and input into the domain model; Obtain the chained Jacobian matrix between each of the key nodes in the domain model; Based on the chained Jacobian matrix, the propagation process of the disturbance between each of the key nodes is determined.

2. The approximate modeling method for information flow in a flight control system oriented towards disturbances according to claim 1, characterized in that, The information flow metamodel includes unit-type elements, information-type elements, and connection-type elements; the unit-type elements include modules, nodes, and transfer functions; the node is the smallest information processing unit executed by the information flow, and the node converts input variables into output variables through the transfer function.

3. The approximate modeling method for information flow in a flight control system oriented towards disturbances according to claim 1, characterized in that, The flight control system is divided into functional modules, which include: Obtain the configuration file of the flight control system; Based on the configuration file, the compilation dependencies between the various modules in the flight control system are determined; Based on the compilation dependencies, the modules of the flight control system are combined and split to obtain the functional modules.

4. The approximate modeling method for information flow in a flight control system oriented towards disturbances according to claim 1, characterized in that, The process of identifying key code blocks from the functional modules includes: The reverse slicing method is used to trace and analyze the data dependencies of the code in the functional module to identify the key code blocks.

5. The approximate modeling method for information flow in a flight control system oriented towards disturbances according to claim 1, characterized in that, The process of obtaining the relevant variables and transfer functions of the key node includes: Obtain the input and output variables of the key nodes; The transfer function is determined based on the input variables and the output variables.

Images