Self-adaptive closed-loop flight control ICD protocol simulation verification system and method oriented to non-ideal environment

Abstract

The invention discloses a non-ideal environment-oriented adaptive closed-loop flight control ICD protocol simulation verification system and method, and belongs to the technical field of software information security analysis, and the non-ideal environment-oriented adaptive closed-loop flight control ICD protocol simulation verification system comprises a basic modeling and simulation module which is used for constructing a flight control system virtual prototype; the non-ideal environment injection module is used for injecting disturbance models of a physical layer, a link layer and a node layer into the virtual bus and the node; the dynamic scene driving module is used for dynamically generating ICD data according to the flight scene data; the intelligent monitoring and analysis module is internally provided with an anomaly detector based on a graph neural network and outputs a structured diagnosis report; and the self-adaptive optimization guide module is internally provided with an optimization strategy engine based on reinforcement learning, generates optimization suggestions and selectively and automatically adjusts configuration to initiate re-verification. The reliability and the research and development efficiency of a flight control system are remarkably improved by simulating a real operation environment in a high-fidelity manner in a design stage.

Description

Technical Field

[0001] This invention relates to the field of software information security analysis technology, and more specifically, to an adaptive closed-loop flight control ICD protocol simulation verification system and method for non-ideal environments. Background Technology

[0002] As the core control unit of aviation equipment, the flight control system's interface control document (ICD) and hardware bus interface matching directly determine the system's reliability, real-time performance, and compatibility, making it a crucial link in ensuring system integration efficiency during aviation equipment development. Currently, with the increasing complexity of aviation equipment functions, flight control systems involve increasingly diverse communication bus types, including various dedicated buses such as TTP, 1553B, CAN, 1394B, and TTE. The communication mechanisms, interface parameters, and data transmission protocols of different buses differ significantly, further increasing the difficulty of verifying the matching between the ICD and hardware bus interfaces.

[0003] In existing technologies, such as the "Flight Control ICD Protocol Simulation Verification System and Flight Control System Based on Virtual Modeling" disclosed in patent CN121056334B, a closed-loop process from ICD parsing to virtual bus simulation verification is achieved by constructing a network topology modeling module, a virtual bus modeling service module, an ICD modeling service module, a processing algorithm modeling service module, and an ICD simulation operation monitoring module. This technology completes the system's network construction by virtually simulating commonly used system network buses during the design phase. Combined with data processing algorithms, it can simulate the business data interaction between system nodes at the system design front end, identify ICD design problems, and complete the compatibility verification of the ICD and hardware bus interface at the design front end, reducing rework waste caused by design mismatch problems.

[0004] However, existing virtual verification environments are usually built on idealized models and have the following technical limitations: First, they fail to fully consider the non-ideal characteristics of physical buses in real operating environments, such as bit flips caused by electromagnetic interference, packet delays or losses caused by bus load conflicts, and signal jitter caused by changes in ambient temperature, resulting in an overly "idealized" simulation environment; Second, ICD data generation is mostly based on preset rules, making it difficult to simulate real-time dynamic data flows that are closely coupled with complex flight scenarios; Third, the verification process is mostly an "open-loop" mode, i.e., verification - problem discovery - manual modification - re-verification, which still needs to be improved in efficiency and lacks the ability to intelligently analyze verification results and guide design optimization. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this invention is to provide an adaptive closed-loop flight control ICD protocol simulation and verification system and method for non-ideal environments, which can significantly improve the reliability and R&D efficiency of the flight control system by simulating the real operating environment with high fidelity during the design stage.

[0006] To solve the above problems, the present invention adopts the following technical solution.

[0007] An adaptive closed-loop flight control ICD protocol simulation verification system and method for non-ideal environments includes: a basic modeling and simulation module for constructing the virtual network topology, virtual bus communication services, Interface Control Document (ICD) data model, and data processing algorithm model of the flight control system, providing a basic operating environment for simulation verification; a non-ideal environment injection module connected to the basic modeling and simulation module for injecting preset or random non-ideal disturbance models into the virtual bus and node data processing during simulation to simulate the non-ideal characteristics of the bus in a real physical environment; and a dynamic scene driving module connected to the basic modeling and simulation module for receiving external flight scene data and converting it into real-time dynamic data generated from ICD data. The system includes: a constraint mechanism to achieve deep coupling between ICD data stream and flight scenario; an intelligent monitoring and analysis module connected to the basic modeling and simulation module and the non-ideal environment injection module, used to monitor simulation operation data in real time, and identify abnormal protocol behavior patterns based on the built-in protocol behavior anomaly detector, outputting a structured diagnostic report containing problem type, occurrence time, associated ICD fields and possible causes; and an adaptive optimization guidance module connected to the intelligent monitoring and analysis module and the basic modeling and simulation module, respectively, used to receive the structured diagnostic report, generate an optimization modification suggestion list based on the built-in optimization strategy engine, and optionally automatically adjust the corresponding configuration in the basic modeling and simulation module to initiate a new round of verification, forming a verification closed loop.

[0008] Furthermore, the non-ideal environment injection module includes: a bus physical layer disturbance model unit, used to simulate signal level distortion, bit flip, or bit stuffing errors caused by electromagnetic interference; a bus data link layer disturbance model unit, used to simulate arbitration delay, data packet collision, retransmission, packet loss, or transmission jitter caused by bus contention; and a node processing layer disturbance model unit, used to simulate data processing delay, task scheduling jitter, memory access errors, or watchdog timeout caused by excessive load on the node's central processing unit.

[0009] Furthermore, the protocol behavior anomaly detector built into the intelligent monitoring and analysis module is an anomaly detection model based on graph neural networks. This model models the flight control system topology as graph data, uses historical simulation data and design specifications as training basis, analyzes the data flow characteristics on the bus in real time, identifies abnormal protocol behavior, and automatically associates the anomaly with the causal relationship between the anomaly and non-ideal environmental disturbances and ICD design defects.

[0010] Furthermore, the optimization strategy engine built into the adaptive optimization guidance module is a reinforcement learning-based optimization strategy engine. This engine encodes the current system configuration parameters into a state space, encodes the system configuration adjustment actions into an action space, and uses the structured diagnostic report output by the intelligent monitoring and analysis module as a negative reward, learning the optimization strategy through offline training.

[0011] A simulation verification method for an adaptive closed-loop flight control ICD protocol for non-ideal environments is also disclosed, comprising the following steps: Step S1: Basic modeling step, constructing the virtual network topology, virtual bus communication service, ICD data model, and data processing algorithm model of the flight control system through the basic modeling and simulation module; Step S2: Non-ideal environment configuration step, configuring a preset or random non-ideal disturbance model through the non-ideal environment injection module, determining the disturbance type, intensity, and time axis arrangement; Step S3: Dynamic scene loading step, receiving and parsing external flight scene data through the dynamic scene driving module, converting it into real-time dynamic constraints for ICD data generation; Step S4: Simulation operation and monitoring step, starting the simulation operation, performing ICD data interaction under non-ideal environment disturbances and dynamic scene constraints, and simultaneously... The intelligent monitoring and analysis module monitors the bus data flow in real time and identifies abnormal protocol behavior patterns based on the built-in protocol behavior anomaly detector, outputting a structured diagnostic report; Step S5: Intelligent Diagnosis Step, based on the structured diagnostic report, analyzes whether the anomaly is caused by an ICD design defect or by non-ideal environmental disturbances, and locates the node, interface, and ICD field where the anomaly occurs; Step S6: Adaptive Optimization Guidance Step, through the optimization strategy engine built into the adaptive optimization guidance module, generates an optimization modification suggestion list based on the structured diagnostic report; Step S7: Closed-Loop Verification Step, adjusts the corresponding configuration in the basic modeling and simulation module according to the optimization modification suggestion list, and reruns the simulation verification under the same or similar non-ideal environmental conditions until the verification passes or the preset number of iterations is reached.

[0012] Furthermore, the configuration of non-ideal disturbance models in step S2 further includes: arranging the combination and intensity of various non-ideal disturbance models according to the time axis through a graphical disturbance scenario configuration interface to generate a reusable disturbance scenario configuration file.

[0013] Furthermore, the dynamic scene loading in step S3 further includes: dynamically adjusting the generation rate, data accuracy, and data range of ICD data based on the flight speed, altitude, and attitude angle information in the flight scene data, and superimposing noise that conforms to the specific sensor characteristics according to a preset sensor noise model.

[0014] Furthermore, the anomaly detection model based on graph neural network in step S4 is a graph attention network model. This model learns the feature distribution of nodes and edges of the flight control system under normal operating conditions, and quickly locates the anomaly propagation path and root node after injection in a non-ideal environment.

[0015] Furthermore, in step S6, the optimization strategy engine based on reinforcement learning is a deep Q-network model. This model calculates the expected benefits of taking different optimization actions based on the current state encoding and diagnostic report, and selects the optimal combination of optimization actions to generate a list of optimization modification suggestions. The optimization actions include one or more combinations of modifying the ICD field length, adding a check field, adjusting the bus scheduling strategy, and adjusting the data transmission cycle.

[0016] A flight control system is also disclosed, wherein the interface control documentation design of the flight control system is verified and optimized using the aforementioned adaptive closed-loop flight control ICD protocol simulation verification system for non-ideal environments.

[0017] Compared with the prior art, the advantages of this invention are:

[0018] Achieving high-fidelity simulation of non-ideal environments: By introducing a non-ideal environment injection module, this invention breaks through the limitations of existing "idealized" simulation technologies, enabling the exposure of compatibility issues between the ICD and hardware under real stress environments during the design phase, and significantly improving the robustness and reliability of the flight control system.

[0019] The invention achieves scene-driven dynamic data generation: Through a dynamic scene-driven module, the static ICD data generation is upgraded to a dynamic generation process that is deeply coupled with the flight scene, thereby qualitatively improving the coverage and realism of simulation verification.

[0020] This invention achieves an intelligent verification closed loop: Through an intelligent monitoring and analysis module and an adaptive optimization guidance module, it upgrades the traditional "open-loop" verification to an "intelligent closed loop" of "discovery-diagnosis-suggestion-optimization". This not only significantly improves the efficiency of problem location and resolution, but also utilizes reinforcement learning technology to provide designers with potentially better design solutions that go beyond the scope of experience, demonstrating significant creativity.

[0021] The automation and intelligence of the verification system have been improved: By introducing artificial intelligence technologies such as graph neural networks and reinforcement learning, the system has the ability to learn, analyze and make decisions autonomously, which greatly reduces human intervention and improves verification efficiency and design quality. Attached Figure Description

[0022] Figure 1 This is a module architecture diagram of the simulation verification system in an embodiment of the present invention;

[0023] Figure 2 This is a schematic diagram of the perturbation model hierarchy of the non-ideal environment injection module in an embodiment of the present invention;

[0024] Figure 3 This is a flowchart illustrating the workflow of the adaptive optimization guidance module in an embodiment of the present invention. Detailed Implementation

[0025] Example 1:

[0026] like Figure 1 As shown, the core architecture of the simulation verification system provided by this invention includes:

[0027] Basic Modeling and Simulation Module 100: This module includes the functions of sub-modules such as network topology modeling, virtual bus modeling, ICD modeling, and processing algorithm modeling from existing patents, and is the cornerstone of the entire simulation. It provides a standardized, configurable virtual prototype of the flight control system for subsequent modules. In terms of implementation, a graphical modeling interface can be used, supporting the construction of nodes and buses through drag-and-drop, and saving the topology configuration in XML format. The virtual bus service uses the DDS distributed bus at its underlying level, shielding the physical bus implementation details, while also supporting the extension of new bus types through dynamic libraries.

[0028] Non-ideal environment injection module 200: This module is deeply coupled with the virtual bus service and data processing service in the basic modeling and simulation module 100. It is not a simple parameter configuration interface, but a dynamic disturbance generator.

[0029] like Figure 2 As shown, it internally maintains a perturbation model library, including:

[0030] Bus physical layer disturbance model unit 201: Based on a random process or a predefined waveform, it modifies the data frame bitstream of the underlying virtual bus to simulate electromagnetic pulse interference. For example, it can randomly flip data bits according to a set bit error rate (BER) or superimpose Gaussian white noise within a specific time period.

[0031] Bus data link layer disturbance model unit 202: Simulates data frame collisions and retransmissions when the bus load is too high by modifying the bus arbitration algorithm or inserting a delay queue. For example, a bus load threshold can be set, and when the simulated data volume exceeds the threshold, data packets are dropped or the transmission delay is increased with a certain probability.

[0032] Node-level perturbation model unit 203: Simulates node computational pressure by hijacking the node data processing thread and injecting random processing delays. For example, after a node receives data, the processing time can be increased according to a set probability distribution to simulate a CPU overload scenario.

[0033] Users can define a "disruption script" through a graphical interface, for example: injecting mild electromagnetic noise at 10-15 seconds of simulation time; and simulating a 90% bus load at 30-35 seconds, triggering packet loss (link layer, 5% packet loss rate). This makes the verification environment closely resemble a real physical integration test environment.

[0034] Dynamic Scene Driving Module 300: This module receives external flight scene data files (such as six-DOF aircraft model data exported from FlightGear or MATLAB). It maps these high-level scene parameters (velocity, altitude, angle of attack, control surface deflection) to changes in the ICD data stream in real time. For example, when the flight scene enters a high-G maneuver turn, this module dynamically adjusts the generation rate and noise level of the attitude angle ICD data packets emitted from the "inertial navigation node" to conform to the sensor characteristics in this physical process, rather than simply sending constant values ​​at a fixed period. In specific implementation, a sensor model library can be pre-established, containing typical noise characteristics (such as Allen variance curves) of sensors such as gyroscopes, accelerometers, and GPS, and appropriate models can be selected and superimposed according to the dynamic parameters in the flight scene.

[0035] Intelligent Monitoring and Analysis Module 400: This module integrates a protocol anomaly detector based on Graph Attention Network (GAT) on top of traditional monitoring functions. It models the flight control system as a graph G=(V,E), where the node set V represents the flight control computer, sensors, actuators, and other devices, and the edge set E represents the bus connections. Each node and edge is associated with a feature vector, such as the node's data transmission rate and error count, and the edge's data frame interval, checksum accuracy, etc. The Graph Attention Network can learn the feature distribution of nodes and edges under normal communication patterns. When a non-ideal environment injection causes anomalies in the ICD protocol, the network model can quickly locate the path of anomaly propagation and provide a diagnostic report. For example, the diagnostic report might be: "The data frame checksum errors received by interface 2 of node C (actuator) have surged. The anomaly propagation path is node A->bus B->node C. The possible cause is data packet corruption due to link layer disturbance, and the ICD protocol does not have a sufficiently strong error recovery mechanism."

[0036] Adaptive Optimized Bootloader 500: Reference Figure 3 The core of this module is a pre-trained Deep Q-Network (DQN) optimization policy engine. Its workflow is as follows:

[0037] Step S501: Receive the diagnostic report from the intelligent monitoring and analysis module 400 and encode it as the current status S_t. The status code includes the current topology, ICD configuration parameters, bus parameters, and the anomaly type and location in the diagnostic report.

[0038] Step S502: The DQN engine calculates the expected reward Q value for taking different optimization actions A_i based on the state S_t and the pre-learned policy π. These actions include, for example, adding a 2-byte CRC checksum to ICD message X, changing the scheduling policy of bus Y from static to dynamic, adding a digital filter to the receiving algorithm of node Z, and adjusting the transmission period of message X. The action space covers multiple dimensions such as ICD design, bus configuration, and algorithm configuration.

[0039] Step S503: The engine selects the top three actions with the highest Q values, generates a list of optimization suggestions, and includes a pre-evaluation score of the system's robustness after each suggestion modification (based on a built-in fast evaluation model, such as a simplified simulator).

[0040] Step S504: Present the suggestion list to the designers via the UI. Designers can choose to adopt the suggestions and modify them manually, or authorize the system to perform "one-click optimization".

[0041] Step S505: If automatic optimization is authorized, the engine will automatically adjust the corresponding configuration parameters in the basic modeling and simulation module 100.

[0042] Step S506: The system automatically reruns the simulation verification under a similar non-ideal environmental configuration.

[0043] Step S507: If the verification passes, the process ends; if it fails, a new monitoring report will drive the optimization process again, forming a closed loop.

[0044] Example 2:

[0045] The DQN engine in the adaptive optimization guidance module 500 needs to be trained offline. The training process is as follows: Build a simulation environment: Use the basic modeling and simulation module 100 to build a typical flight control system model, which includes various bus types and nodes.

[0046] Define the action space: determine all possible optimization actions, such as modifying the ICD field length, adding verification, adjusting the bus speed, changing the scheduling algorithm, etc.

[0047] Define the state space: Encode the system's configuration parameters (ICD format, bus parameters, topology connections) and the features of the diagnostic reports output by the intelligent monitoring module into state vectors.

[0048] Define a reward function: The reward function is designed to reflect the severity of the diagnostic reports negatively. If no abnormal diagnostic reports are generated after a simulation run, a high positive reward is given; if a severe abnormal report is generated, a negative reward (penalty) is given. The reward can also consider optimizing the cost, such as providing additional rewards for actions with small modifications that do not affect real-time performance.

[0049] Training: Connect the DQN agent to the simulation environment. The agent randomly selects actions and observes new states and rewards. Through experience replay, the Q-network parameters are continuously updated until convergence. The trained DQN engine can quickly provide the optimal action combination when faced with new diagnostic reports.

[0050] Through the collaborative work of the above modules, this invention not only provides a more realistic simulation verification environment, but also builds an intelligent verification platform that can learn and optimize itself, greatly improving the efficiency and reliability of flight control ICD design.

[0051] Example 3:

[0052] The specific implementation steps of the method of the present invention are as follows:

[0053] Step S1: Construct a system model using the basic modeling and simulation module, including a flight control computer node, a GPS sensor node, an inertial measurement unit node, a servo actuator node, and a CAN bus connecting each node.

[0054] Step S2: Configure interference scenarios through the non-ideal environment injection module: normal operation for 0-5 seconds in simulation; physical layer disturbance (bit error rate 10^-5) injected for 5-10 seconds; normal operation restored for 10-15 seconds; link layer disturbance (bus load 80%, packet loss rate 2%) injected for 15-20 seconds.

[0055] Step S3: Load scenario data for a test flight mission through the dynamic scenario-driven module, including flight parameters for stages such as takeoff, climb, cruise, maneuvering turns, and landing.

[0056] Step S4: Start the simulation. During the maneuvering and turning phase, the attitude data update rate increases due to the flight scenario, and the added link layer disturbances cause some data packets to be lost.

[0057] Step S5: The intelligent monitoring and analysis module detected an anomaly based on the graph attention network and located a packet loss problem in the data flow from the inertial measurement unit node to the flight control computer node. The diagnostic report pointed out: "The packet loss rate exceeds the threshold, which may be caused by excessive bus load and excessively short ICD data cycle setting coupling."

[0058] Step S6: The DQN engine of the adaptive optimization guidance module generates three optimization suggestions based on the diagnostic report: Suggestion A - Adjust the attitude data transmission period from 10ms to 20ms; Suggestion B - Add a 2-byte CRC check to the attitude data packet and enable the retransmission mechanism; Suggestion C - Send critical data and non-critical data in a staggered time-sharing manner.

[0059] Step S7: The designer selects to adopt the combination of suggestions A and C. The system automatically adjusts the configuration and reruns the verification. In the second round of simulation under the same interference scenario, the packet loss rate is reduced to an acceptable range, and the verification is successful.

Claims

1. An adaptive closed-loop flight control ICD protocol simulation and verification system for non-ideal environments, characterized in that, include: The basic modeling and simulation module is used to construct the virtual network topology, virtual bus communication services, interface control document (ICD) data model, and data processing algorithm model of the flight control system, providing a basic operating environment for simulation verification. The non-ideal environment injection module, connected to the basic modeling and simulation module, is used to inject preset or random non-ideal disturbance models into the virtual bus and node data processing during simulation to simulate the non-ideal characteristics of the bus in the real physical environment. The dynamic scene driving module, connected to the basic modeling and simulation module, is used to receive external flight scene data and convert it into real-time dynamic constraints for ICD data generation, thereby achieving deep coupling between ICD data stream and flight scene. The intelligent monitoring and analysis module, connected to the basic modeling and simulation module and the non-ideal environment injection module, is used to monitor simulation operation data in real time, identify abnormal protocol behavior patterns based on the built-in protocol behavior anomaly detector, and output a structured diagnostic report containing the problem type, occurrence time, associated ICD fields and possible causes. as well as The adaptive optimization guidance module is connected to the intelligent monitoring and analysis module and the basic modeling and simulation module, respectively. It is used to receive the structured diagnostic report, generate a list of optimization modification suggestions based on the built-in optimization strategy engine, and optionally automatically adjust the corresponding configuration in the basic modeling and simulation module to initiate a new round of verification, forming a verification closed loop.

2. The adaptive closed-loop flight control ICD protocol simulation and verification system for non-ideal environments according to claim 1, characterized in that, The non-ideal environment injection module includes: a bus physical layer disturbance model unit, used to simulate signal level distortion, bit flip, or bit stuffing errors caused by electromagnetic interference; a bus data link layer disturbance model unit, used to simulate arbitration delay, data packet collision, retransmission, packet loss, or transmission jitter caused by bus contention; and a node processing layer disturbance model unit, used to simulate data processing delay, task scheduling jitter, memory access errors, or watchdog timeout caused by excessive load on the node's central processing unit.

3. The adaptive closed-loop flight control ICD protocol simulation and verification system for non-ideal environments according to claim 1, characterized in that: The built-in protocol behavior anomaly detector of the intelligent monitoring and analysis module is an anomaly detection model based on graph neural networks. This model models the flight control system topology as graph data, uses historical simulation data and design specifications as training basis, analyzes the data flow characteristics on the bus in real time, identifies abnormal protocol behavior, and automatically associates the anomaly with the causal relationship between the anomaly and non-ideal environmental disturbances and ICD design defects.

4. The adaptive closed-loop flight control ICD protocol simulation and verification system for non-ideal environments according to claim 1, characterized in that: The adaptive optimization guidance module has a built-in optimization strategy engine based on reinforcement learning. This engine encodes the current system configuration parameters into a state space, encodes the system configuration adjustment actions into an action space, and uses the structured diagnostic report output by the intelligent monitoring and analysis module as a negative reward. It learns optimization strategies through offline training.

5. A simulation verification method for an adaptive closed-loop flight control ICD protocol oriented towards a non-ideal environment based on the system described in any one of claims 1 to 4, characterized in that, Includes the following steps: Step S1: Basic modeling step, constructing the virtual network topology, virtual bus communication service, ICD data model and data processing algorithm model of the flight control system through the basic modeling and simulation module; Step S2: Non-ideal environment configuration step, through the non-ideal environment injection module, configure the preset or random non-ideal disturbance model, and determine the disturbance type, intensity and time axis arrangement; Step S3: Dynamic scene loading step, which receives and parses external flight scene data through the dynamic scene driving module and converts it into real-time dynamic constraints generated from ICD data; Step S4: Simulation operation and monitoring steps. Start the simulation operation and perform ICD data interaction under non-ideal environmental disturbances and dynamic scene constraints. At the same time, monitor the bus data flow in real time through the intelligent monitoring and analysis module, and identify abnormal protocol behavior patterns based on the built-in protocol behavior anomaly detector, and output a structured diagnostic report. Step S5: Intelligent diagnosis step, based on the structured diagnosis report, analyze whether the anomaly is caused by ICD design defects or by non-ideal environmental disturbances, and locate the node, interface and ICD field where the anomaly occurred; Step S6: Adaptive optimization guidance step, which uses the optimization strategy engine built into the adaptive optimization guidance module to generate a list of optimization modification suggestions based on the structured diagnostic report; Step S7: Closed-loop verification step. Adjust the corresponding configuration in the basic modeling and simulation module according to the optimization and modification suggestion list, and rerun the simulation verification under the same or similar non-ideal environmental conditions until the verification is passed or the preset number of iterations is reached.

6. The simulation and verification method for adaptive closed-loop flight control ICD protocol in non-ideal environments according to claim 5, characterized in that: The configuration of non-ideal disturbance models in step S2 further includes: arranging the combination and intensity of various non-ideal disturbance models according to the time axis through a graphical disturbance scenario configuration interface to generate a reusable disturbance scenario configuration file.

7. The simulation and verification method for adaptive closed-loop flight control ICD protocol in non-ideal environments according to claim 5, characterized in that: The dynamic scene loading in step S3 further includes: dynamically adjusting the generation rate, data accuracy, and data range of ICD data based on the flight speed, altitude, and attitude angle information in the flight scene data, and superimposing noise that conforms to the specific sensor characteristics according to the preset sensor noise model.

8. The simulation and verification method for adaptive closed-loop flight control ICD protocol in non-ideal environments according to claim 5, characterized in that: In step S4, the anomaly detection model based on graph neural network is a graph attention network model. This model learns the feature distribution of nodes and edges of the flight control system under normal operating conditions, and quickly locates the anomaly propagation path and root node after injection in a non-ideal environment.

9. The simulation and verification method for adaptive closed-loop flight control ICD protocol in non-ideal environments according to claim 5, characterized in that: In step S6, the optimization strategy engine based on reinforcement learning is a deep Q-network model. This model calculates the expected benefits of taking different optimization actions based on the current state encoding and diagnostic report, and selects the optimal combination of optimization actions to generate a list of optimization modification suggestions. The optimization actions include one or more combinations of modifying the ICD field length, adding a check field, adjusting the bus scheduling strategy, and adjusting the data transmission cycle.

10. A flight control system, characterized in that, The interface control documentation design of the flight control system is verified and optimized using the adaptive closed-loop flight control ICD protocol simulation verification system for non-ideal environments as described in any one of claims 5 to 9.

Images