Method and system for building flight control computer simulation test system

Abstract

The invention provides a method and system for building a flight control computer simulation test system, and belongs to the technical field of aircrafts, and the method comprises the steps: building a structured simulation test knowledge base of an aircraft; a flight control semi-physical simulation test environment and an intelligent agent for operating and sensing the flight control semi-physical simulation test environment are established, and the intelligent agent is formed through training of a structured simulation test knowledge base of the aircraft; building a sensing layer, a decision layer, a memory layer and an execution layer for the intelligent agent to form a flight control simulation intelligent agent; and physically connecting the flight control semi-physical simulation test environment with the flight control simulation agent to form a flight control computer simulation test system. According to the invention, the flight control simulation intelligent agent is used for carrying out test design, test execution control and result analysis on a flight control semi-physical simulation test environment, so that the purpose of autonomously completing a simulation test without relying on manpower is achieved, the influence of introduction of human errors is avoided, and the reliability of a simulation test result is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of aircraft, and relates to the technology of combining the flight control computer of an aircraft with an intelligent agent for simulation, specifically to a method and system for building a flight control computer simulation test system. Background Art

[0002] The flight control computer is a key component of an aircraft, especially as the core module of an unmanned aircraft. The quality of its performance directly affects the flight safety and mission execution ability of the aircraft. Due to the high cost of aircraft design and the possible risks in the test phase, the simulation test has become an indispensable test verification link in the aircraft design process. The simulation test of the flight control computer is to simulate the real flight environment of the aircraft, test and verify the software and hardware of the flight control computer, so as to improve the R & D design efficiency, effectively shorten the design cycle, and reduce the cost and risk of aircraft design.

[0003] However, with the increasing diversification of the functions of the flight control computer, the types of simulation test data and the requirements of simulation projects of the flight control computer simulation test system are also becoming more and more diverse. At the same time, the reliability requirements for the test results generated by the flight control computer simulation test system are also getting higher and higher. The processes such as test design, test execution, and result analysis in the flight control computer simulation test process basically need to be completed manually, which requires a large amount of manpower in the flight control computer simulation test. When processing the simulation test data, it is easily interfered by human factors, resulting in large errors in the process of processing the simulation test data and affecting the reliability of the simulation test results. Summary of the Invention

[0004] In view of the technical problems described in the above background art, that is, the existing flight control computer simulation test system requires a large amount of manpower in the simulation test process, is easily interfered by artificial factors when processing the simulation test data, resulting in large errors in the process of processing the simulation test data and affecting the reliability of the simulation test results. To solve this technical problem, the present invention proposes a method and system for building a flight control computer simulation test system.

[0005] The present invention introduces an intelligent agent into the flight control computer simulation test, trains the intelligent agent using the structured simulation test knowledge base of the aircraft, and completes the test design, test execution, and result analysis of the flight control computer simulation test through the intelligent agent, so as to achieve the purpose of autonomously completing the simulation test without relying on manpower, avoiding the artificial errors introduced in the flight control computer simulation test process, and improving the reliability of the simulation test results.

[0006] To solve the above technical problems, the present invention adopts the following technical solutions: A method for building a flight control computer simulation test system of the present invention includes: Constructing a structured simulation test knowledge base for the aircraft; Building a flight control semi-physical simulation test environment and an agent for operating and perceiving the flight control semi-physical simulation test environment, where the agent is trained through the structured simulation test knowledge base of the aircraft; Building a perception layer, a decision-making layer, a memory layer, and an execution layer for the agent to form a flight control simulation agent; among them, the decision-making layer of the agent uses the scoring function of the decision-making path and the optimization function of the decision-making path to screen the decision-making path, and then uses the dynamic pruning strategy to eliminate the decision-making paths that do not meet the dynamic pruning conditions after screening. The dynamic pruning conditions include the decision-making path threshold and the number of executions; the decision-making path refers to the flight path of the aircraft in the simulation test; Physically connecting the flight control semi-physical simulation test environment and the flight control simulation agent to form a flight control computer simulation test system.

[0007] Further limited, the screening using the scoring function of the decision-making path and the optimization function of the decision-making path specifically includes: Determining the total score of each decision-making path according to the node actions of each decision-making path and the rewards obtained after the execution of the node actions; Constructing a cost function, normalizing the cost function to form an optimization function of the decision-making path; Judging the total scores of all decision-making paths according to the optimization function of the decision-making path, screening the decision-making paths, and forming the screened decision-making paths.

[0008] Further limited, the eliminating of the decision-making paths that do not meet the dynamic pruning conditions after screening using the dynamic pruning strategy specifically includes: Constructing an effectiveness pruning formula for the decision-making path, dynamically evaluating the screened decision-making paths according to the effectiveness pruning formula of the decision-making path, and eliminating the decision-making paths that do not meet the dynamic pruning conditions in the screened decision-making paths according to the dynamic evaluation results.

[0009] Further limited, the constructing of the structured simulation test knowledge base for the aircraft specifically includes: Obtaining the simulation test knowledge base of the aircraft; Using data sorting algorithms, data cleaning strategies, data tokenization algorithms, and standardization processing algorithms to classify, clean noise, tokenize data, and label data for the simulation test knowledge base of the aircraft in sequence to form a structured simulation test knowledge base for the aircraft.

[0010] Further limited, the scoring function of the decision-making path is: In the formula, , is the i-th decision path of the aircraft in the simulation test; is the t-th node action on the i-th decision path, t≥1; T is the total number of node actions on the i-th decision path; is the decision factor, r t is the reward score of the t-th node action; is the penalty weight parameter; is the optimization function of the i-th decision path of the aircraft in the simulation test; The validity pruning formula is: And In the formula, is the scoring threshold of the decision path; S i is the total number of successful executions of the node actions by the aircraft in the i-th decision path in the simulation test; n i is the total number of executions of the node actions by the aircraft in the i-th decision path in the simulation test; N min is the minimum effective execution times.

[0011] Further defined, the method for building the flight control computer simulation test system further includes establishing a data dictionary, and associating the perception layer of the intelligent agent and the decision-making layer of the intelligent agent through the data dictionary, so that the decision-making layer of the intelligent agent can call the information perceived in the perception layer of the intelligent agent.

[0012] Further defined, the method for building the flight control computer simulation test system further includes configuring a standard API call interface, and the flight control semi-physical simulation test environment and the flight control simulation intelligent agent are physically connected through the standard API call interface.

[0013] Further defined, the simulation test knowledge base of the aircraft includes the initial scenario information of the flight control computer, the purpose information of the simulation test of the flight control computer, the method information of the simulation test of the flight control computer, the process information of the simulation test of the flight control computer, the data analysis method of the simulation test of the flight control computer, the judgment standard of the simulation test result of the flight control computer, the sensor data of the flight control computer, the simulation case analysis of the flight control computer, the simulation test outline of the flight control computer, and the simulation test report of the flight control computer.

[0014] Based on the above method for building a flight control computer simulation test system, a building system of the flight control computer simulation test system includes: A construction module: used to construct a structured simulation test knowledge base of the aircraft; The first construction module: It is used to construct the flight control hardware-in-the-loop simulation test environment and the intelligent agent for operating and perceiving the flight control hardware-in-the-loop simulation test environment. Among them, the intelligent agent is trained through the structured simulation test knowledge base of the aircraft. The second construction module: It is used to construct a perception layer, a decision-making layer, a memory layer, and an execution layer for the intelligent agent to form a flight control simulation intelligent agent. Among them, the decision-making layer of the intelligent agent uses the scoring function of the decision-making path and the optimization function of the decision-making path to screen the decision-making path, and then uses the dynamic pruning strategy to eliminate the decision-making paths that do not meet the dynamic pruning conditions after screening. The dynamic pruning conditions include the decision-making path threshold and the number of executions; the decision-making path refers to the flight path of the aircraft in the simulation test. And the system formation module: It is used to physically connect the flight control hardware-in-the-loop simulation test environment with the flight control simulation intelligent agent to form a flight control computer simulation test system.

[0015] The present invention proposes a memory storing program files, and the program files are executed to implement program instructions formed by the method for constructing the flight control computer simulation test system as described above.

[0016] The present invention proposes an electronic device, including a processor and a memory coupled to each other. Among them, The memory: It is used to store program instructions formed by the method for constructing the flight control computer simulation test system as described above. The processor: It is used to execute the program instructions stored in the memory.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. In the method for constructing the flight control computer simulation test system of the present invention, an intelligent agent is introduced and trained through the structured simulation test knowledge base of the aircraft to form a flight control simulation intelligent agent; the flight control hardware-in-the-loop simulation test environment is physically connected with the flight control simulation intelligent agent. The present invention uses the flight control simulation intelligent agent to conduct test design, test execution control, and result analysis for the flight control hardware-in-the-loop simulation test environment, so as to achieve the purpose of autonomously completing the simulation test without relying on human power, avoiding the human errors introduced in the flight control computer simulation test process, and improving the reliability of the simulation test results.

[0018] 2. In the present invention, the decision-making layer of the intelligent agent screens decision-making paths by using the scoring function of the decision-making path and the optimization function of the decision-making path, and then uses the dynamic pruning strategy to eliminate the decision-making paths that do not meet the dynamic pruning conditions after screening; a two-stage optimization mechanism is formed through screening and the dynamic pruning strategy to ensure that the remaining decision-making paths have high execution efficiency, stability, and task completion rate, thereby enabling the intelligent agent to have better planning and reasoning capabilities in the simulation test, and improving the intelligent level and decision-making response quality of the simulation test process.

[0019] 3. The method for building a flight control computer simulation test system of the present invention further includes configuring a standard API call interface, and physically connecting the flight control semi-physical simulation test environment and the flight control simulation intelligent agent through the standard API call interface, so that the information of the flight control semi-physical simulation test environment can be operated and perceived by the flight control simulation intelligent agent. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a schematic diagram of the method for building a flight control computer simulation test system of the present invention; Figure 2 is a schematic diagram of the system for building a flight control computer simulation test system of the present invention; Figure 3 is a schematic diagram of the training process of the intelligent agent. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The technical solution of the present invention will be further explained below in conjunction with the drawings and embodiments, but the present invention is not limited to the following embodiments.

[0022] Essentially, the flight control computer simulation test is to design a large number of tests for different subjects of the flight control computer to verify its design, functions, algorithm logic, strategies, and performance and other indicators. Therefore, the processes of test design, test execution, and result analysis are universal. The present invention proposes a method, system, memory, and electronic device for building a flight control computer simulation test system based on the underlying logic, test process, and data analysis methods relied on by the flight control computer simulation test.

[0023] See Figure 1 , the present invention proposes a method for building a flight control computer simulation test system, including: S1: Construct a structured simulation test knowledge base for the aircraft; the structured simulation test knowledge base for the aircraft is formed based on the simulation test knowledge base of the aircraft, and the simulation test knowledge base of the aircraft includes initial scenario information of the flight control computer, simulation test purpose information of the flight control computer, simulation test method information of the flight control computer, simulation test process information of the flight control computer, simulation test data analysis method of the flight control computer, simulation test result determination criteria of the flight control computer, sensor data of the flight control computer, simulation case analysis of the flight control computer, simulation test outline of the flight control computer, simulation test system user manual of the flight control computer, and simulation test reports of the flight control computer, etc.

[0024] Preferably, the present invention further includes building a data processing framework, which includes a data sorting algorithm, a data cleaning strategy, a data word segmentation algorithm, and a standardization processing algorithm. The simulation test knowledge base of the aircraft is sorted, noise-cleaned, data word-segmented, and data labeled in sequence through the data processing framework, so as to convert the simulation test information into JSON structured data. The JSON structured data is used as the structured simulation test knowledge base of the aircraft, and the intelligent agent is trained using the structured simulation test knowledge base of the aircraft, which is convenient for the intelligent agent to understand and use.

[0025] S2: Build a flight control semi-physical simulation test environment and an intelligent agent for operating and perceiving the flight control semi-physical simulation test environment, where the intelligent agent is trained through the structured simulation test knowledge base of the aircraft.

[0026] In the present invention, the language model structure of the intelligent agent can be selected from an open-source large model, such as Tongyi Qianwen, iFlytek Spark, Wenxin Yiyan, Zhipu Qingyan, and Tencent Hunyuan, etc. Preferably, the language model structure of the intelligent agent in the present invention is Wenxin Yiyan 4.5. The present invention selects the Transformer framework in Wenxin Yiyan 4.5 for training through the structured simulation test knowledge base of the aircraft, so that the Transformer framework predicts the next word in the sentence or fills in the missing word in the sentence, etc. The specific training process is as Figure 3 shown, and the process includes constructing the structured simulation test knowledge base of the aircraft, inputting questions, starting training, similarity matching, relevance extraction, word segmentation and completion, and training evaluation, so as to form a flight simulation intelligent agent.

[0027] In the present invention, the flight control semi-physical simulation test environment is composed of a flight control computer, a real-time simulator, a dynamics model, a sensor model, simulation management software, simulation scheduling software, and various buses. Specifically, building the flight control semi-physical simulation test environment is a well-known technology to those skilled in the art, and will not be elaborated in the present invention.

[0028] S3: Build a perception layer, a decision-making layer, a memory layer, and an execution layer for the agent to form a flight control simulation agent. Among them, the decision-making layer of the agent uses the scoring function of the decision-making path and the optimization function of the decision-making path to screen the decision-making paths, and then uses the dynamic pruning strategy to eliminate the decision-making paths that do not meet the dynamic pruning conditions after screening. The dynamic pruning conditions include the decision-making path threshold and the number of execution times. The decision-making path refers to the flight path of the aircraft in the simulation test experiment.

[0029] S4: Physically connect the flight control semi-physical simulation test environment with the flight control simulation agent to form a flight control computer simulation test system.

[0030] The method for building the flight control computer simulation test system of the present invention further includes configuring a standard API call interface. The flight control semi-physical simulation test environment and the flight control simulation agent are physically connected through the standard API call interface. The standard API call interface is oriented to the flight control simulation agent and provides tools for the flight control simulation agent to operate and perceive the flight control semi-physical simulation test environment.

[0031] The specific process of configuring the standard API call interface is as follows: Unitize various control logics in the flight control semi-physical simulation test environment to generate control units, and each control unit can describe a single function; Standardize each control unit to form a standard API call interface to ensure that each standard API call interface can be registered by the flight control simulation agent, and the flight control simulation agent does not need to do too much thinking and learning when calling the standard API call interface. Among them, the standard API call interface includes a simulation scenario initialization API, a simulation scheduling API, a fault injection API, a model operation API, and a device control API; Standardize the data query structure of the flight control semi-physical simulation test environment, convert natural language data into structured data, and form a knowledge base query interface and a data query interface. The structure of this structured data is the same as the structure of the agent's parsed data, ensuring that the queried data can be efficiently understood.

[0032] In the present invention, a perception layer is built for the agent to enable it to have the ability to obtain data. Specifically, it includes: Summarize the knowledge base query interface and the data query interface, establish a data dictionary, and associate the perception layer of the agent with the decision-making layer of the agent through the data dictionary, so that the decision-making layer of the agent can call the information perceived in the perception layer of the agent. The perception layer of the agent uses a data search algorithm, and the data search algorithm of the present invention is preferably a depth-first algorithm, which matches the decision-making layer of the agent, can ensure that the logic of searching data is consistent with the logic of the decision-making layer of the agent, and is a chain logic.

[0033] In the present invention, building a decision-making layer for an agent includes screening decision-making paths by using a scoring function of the decision-making path and an optimization function of the decision-making path, and then using a dynamic pruning strategy to eliminate the decision-making paths that do not meet the dynamic pruning conditions after screening. The dynamic pruning conditions include a decision-making path threshold and the number of executions.

[0034] Among them, the specific process of screening by using the scoring function of the decision-making path and the optimization function of the decision-making path includes: Determining the total score of each decision-making path according to the node actions of each decision-making path and the rewards obtained after the execution of the node actions; Scoring function of the decision-making path is: In the formula, , is the i-th decision-making path of the aircraft in the simulation test. Each decision-making path consists of a set of sequentially executed node actions; is the t-th node action on the i-th decision-making path, t≥1, which corresponds to the specific operation instructions executed by the flight control computer in the simulation test. For example: initializing the navigation module, starting flight state acquisition, injecting inertial navigation faults or switching control laws, etc.; T is the total number of node actions on the i-th decision-making path, that is, the total number of operation steps in the simulation test; is a decision-making factor, which is used to weighted accumulate the reward scores obtained by each node action, representing the importance of the medium- and long-term actions in the decision-making path. Its value is usually set in the range of [0,1], mainly considering the strategy stability in the aircraft control system; r t is the reward score of the t-th node action. This value is directly related to the accuracy, completion degree of the execution of the node action and its positive contribution to the mission goal. For example, when the flight control computer successfully identifies an inertial navigation anomaly and switches to the backup navigation, r t is a positive value (such as +1.0); conversely, if the execution fails or there is no response, then r t is a negative value (such as -1.0); is a penalty weight parameter; in the aircraft control system, decision-making paths with long execution time, high resource consumption or high task complexity will bear a higher penalty weight; is the optimization function of the i-th decision-making path of the aircraft in the simulation test, which is used to measure the comprehensive cost required for the path to complete the test task, specifically including computing resource consumption (such as CPU occupancy rate), the number of sensor reading calls, control response delay, and the total execution time of the test scenario, etc.; Penalty weight parameter is used to balance the rewards and costs of each decision-making path. Its calculation formula is: In the formula, is the average reward of the historical decision path; is the average cost of the historical decision path.

[0035] Construct a cost function, normalize the cost function, and form an optimization function for the decision path; The cost function is used to evaluate the cost required for a single decision path to be executed, including time consumption and resource consumption. The specific calculation formula is: In the formula, is the cost function of the i-th decision path of the aircraft in the simulation test; t is the serial number of the node action on the i-th decision path; T is the total number of node actions on the i-th decision path; is the adjustment coefficient of time consumption, generally taking a value of 0.6; is the adjustment coefficient of resource consumption, generally taking a value of 0.4; is the time consumption of the t-th node action; is the resource consumption of the t-th node action of the aircraft in the simulation test.

[0036] Normalizing the cost function is to solve the problem of scoring imbalance caused by inconsistent cost scales of different decision paths, and normalize to a reasonable range to obtain the optimization function of the decision path. The specific formula is as follows: In the formula, is the optimization function of the i-th decision path of the aircraft in the simulation test; is the cost function of the i-th decision path of the aircraft in the simulation test; is the known minimum cost of the historical decision path; is the known maximum cost of the historical decision path.

[0037] Evaluate the total score of all decision paths according to the optimization function of the decision path, screen the decision paths, and form the screened decision paths.

[0038] To avoid the influence of the length of the decision path (the i-th decision path) on the score, divide the score function of the decision path by the total number T of node actions on the i-th decision path to obtain the average score of the node actions on the i-th decision path , and use this to judge whether this decision path is a valid path. The formula is: To prevent the over-expansion of the decision layer generated during training and maintain high efficiency, a dynamic pruning strategy is used to eliminate the decision paths that do not meet the dynamic pruning conditions after screening. Specifically, it includes: constructing an effectiveness pruning formula for decision paths, dynamically evaluating the screened decision paths according to the effectiveness pruning formula for decision paths, and eliminating the decision paths that do not meet the dynamic pruning conditions in the screened decision paths based on the dynamic evaluation results.

[0039] The effectiveness pruning formula is: And In the formula, is the scoring threshold of the decision path, which is used to set the lower limit of the score for determining whether a decision path is effective, and reflects the tolerance requirements of the task (node action) for the stability and effectiveness of the decision path in the simulation test. In the simulation test, if the average score of a certain decision path is lower than this scoring threshold, it is regarded as a poor decision path and will be pruned to avoid interfering or misleading the aircraft control system; S i is the total number of successful executions of the node action of the aircraft in the i-th decision path in the simulation test, that is, the number of times the decision path completes the established node action and outputs an acceptable result. Usually, the system state being stable, the flight control response being correct, or the expected data being successfully output is used as the judgment criterion; n i is the total number of executions of the node action of the aircraft in the i-th decision path in the simulation test, including various situations such as success, failure, and interruption; combined with S i can be used to calculate the execution success rate of the decision path; N min is the minimum effective execution number, which is an important parameter for measuring the statistical reliability of the decision path evaluation result. If the execution number of a certain decision path is less than N min , even if the current score is high, it may be regarded as an unstable decision path due to insufficient samples and thus be eliminated by the dynamic pruning strategy. This value is generally set according to the simulation test cycle and task complexity. For example, setting it to 100 times can ensure that the path has basic statistical significance.

[0040] In the present invention, in the decision-making layer of the intelligent agent, a scoring function of the decision-making path is used to quantitatively evaluate and preferentially sort multiple feasible decision-making paths. First, the total score of each decision-making path is calculated based on the rewards and costs of the node actions in the simulation test experiment task. An optimization function of the decision-making path is formed by normalizing the costs to adjust the evaluation deviation caused by inconsistent dimensions between different decision-making paths, making the scoring mechanism more objective and consistent. The decision-making layer of the intelligent agent screens according to the total score, eliminating inefficient or invalid decision-making paths. Further, a dynamic pruning strategy is adopted to perform secondary compression on the selected decision-making paths. Discrimination is made based on the known minimum cost, known maximum cost, average reward, and average cost of the historical decision-making paths of the decision-making path, and those decision-making paths with poor stability and low experimental value are eliminated. A two-stage optimization mechanism is formed through screening and dynamic pruning strategies to ensure that the remaining decision-making paths have high execution efficiency, stability, and task completion degree, so as to enable the intelligent agent to have better planning and reasoning capabilities in the simulation test, and improve the intelligent level and decision-making response quality of the simulation test process.

[0041] In the present invention, a memory layer is built for the intelligent agent, including long-term memory and short-term conversation, specifically including: building a short-term conversation based on the context management mode to store the initial scene settings, experiment process, experiment data, and experiment results of a single simulation experiment in the flight control computer simulation; selecting a vector database to build long-term memory to store basic knowledge, decision-making information, training experience, scene information, skill information, and various mapping relationships in the flight control computer simulation.

[0042] In the present invention, an execution layer is built for the intelligent agent to construct its ability to execute specific actions. The standard API call interface is registered to the intelligent agent and identified so that it can understand the function of the standard API call interface and call it during the task execution process; a data analysis algorithm library is established to provide an algorithm basis for the execution layer. The data analysis algorithm library includes simulation data preprocessing algorithms, data statistical analysis algorithms, and specific algorithms commonly used in simulation. Among them, the simulation data preprocessing algorithms include data interception, bad point elimination, and smoothing processing, etc.; the data statistical analysis algorithms include mean, variance, RMS value, and test algorithms, etc.; the specific algorithms commonly used in simulation include spectrum analysis, filter analysis, and Fourier transform, etc.

[0043] The following takes the process of independently completing the simulation test of the inertial navigation redundancy test as an example to specifically illustrate the construction method of the flight control computer simulation test system of the present invention.

[0044] 1) The user inputs the user requirement "test whether the redundancy determination logic of the flight control computer for inertial navigation is correct" to the flight control simulation intelligent agent; 2) After receiving the user's requirements, the flight control simulation agent conducts understanding. The specific understanding process is as follows: ① The flight control simulation agent extracts the core of the target, including inertial navigation: using accelerometers and gyroscopes to determine the position, velocity, and attitude of an object without relying on external signals; redundancy determination: in the case of redundant sensors, determining whether the sensors are faulty or ineffective and taking corresponding strategies. ② The flight control simulation agent clarifies the test objectives and verifies whether the redundancy determination logic in the flight control computer is correct. Possible tests include detections under normal conditions, single-point failures, multi-point failures, progressive failures, and situations of sensor noise and interference.

[0045] 3) The decision-making layer of the agent splits the task and designs the test steps. ① The test steps include: scene initialization of the flight control computer, simulation scheduling of the flight control computer, data acquisition of the flight control computer, fault injection into the flight control computer, result analysis of the flight control computer, and report generation of the flight control computer, thereby conducting test design. ② According to the evaluation of the decision-making layer of the agent, determine whether the test steps are executable. If not, external tools need to be called and the decision-making layer of the agent is updated simultaneously, thereby conducting test execution. ③ Summarize the test results, analyze the performance of the redundancy determination logic under different conditions. If there are problems, optimize and adjust them specifically. The entire process must ensure the accuracy and reliability of the data, and it may be necessary to repeat multiple tests to confirm the stability of the results, thereby conducting result analysis.

[0046] 4) The flight control simulation agent executes the test steps and maps the test tasks to specific standard API call interfaces through a data dictionary to complete the test steps. Specifically: ① Call the simulation scene initialization API to complete the simulation initialization task. ② Call the simulation scheduling API to control the simulation computer and dynamic model in the flight control hardware-in-the-loop simulation test environment to start the simulation test. ③ Call the data query interface to collect the position and attitude data of the primary inertial navigation, backup inertial navigation, flight control, etc. ④ Call the fault injection API to introduce errors to the primary inertial navigation, such as biases, increased noise, or complete failure. ⑤ Repeat the above steps ① - ⑤ to combine multiple different fault scenarios.

[0047] 5) The flight control simulation agent records various test data from the start to the end of the simulation test through short-term memory, including: test data such as decision execution status, time consumption, and resource consumption.

[0048] 6) The flight control simulation agent analyzes the test data to determine whether the redundancy determination logic of the flight control computer is correct, specifically including: ① During the normal simulation process, monitor the primary inertial navigation position and attitude information of the flight control computer to determine that the redundancy determination logic of the flight control computer for inertial navigation is correct under normal circumstances; ② After injecting a fault into the primary inertial navigation of the flight control computer, monitor the flight control computer and find that it uses the backup inertial navigation position and attitude information to determine that the redundancy determination logic of the flight control computer for inertial navigation is correct in the case of a primary inertial navigation fault; ③ Similarly, multi-point fault testing, progressive fault testing (signal gradually weakening), noise and interference testing, etc. can be carried out to verify whether its determination logic is correct.

[0049] 7) The flight control simulation agent calls the simulation test report template to generate a test report, specifically including: ① The execution layer of the agent obtains the simulation test report template from the database responsible for long-term memory; ② The execution layer of the agent fills in the data such as test steps, test scenarios, test results and conclusions into the template to generate a test report.

[0050] The process of building the decision-making layer of the agent is described below with specific data: Score the generated decision-making path according to the above-mentioned independently completed simulation test inertial navigation redundancy test process, and set the decision-making factor to 0.9, and the reward score r of the action of the t-th node t See Table 1.

[0051] Table 1: Reward scores of node actions on the decision-making path Execution result Reward score Node action executed successfully + data complete +1.0 Node action partially successful / data incomplete +0.5 Node action execution failed / no response -1.0 Combined with the historical data of the flight control computer simulation test, the average reward of the historical decision-making path is 0.75, and the average cost of the historical decision-making path is 0.6. Therefore, the penalty weight parameter is 0.625.

[0052] Combined with the characteristics of the flight control computer simulation test, considering that the proportion of time consumption is higher than the proportion of resource consumption, therefore, set ; Record the execution situation of the decision-making path as shown in Table 2.

[0053] Table 2: Execution situation of the decision-making path Serial number Execution steps Reward score Node action duration Node action resource consumption 1 Call the model initialization and simulation scenario initialization APIs to complete the simulation initialization task 1 10 21 2 Call the start simulation API to control the simulation computer, dynamic model, etc. to start the simulation 1 12 17 3 Call the data acquisition interface to collect the position and attitude data of the primary inertial navigation, backup inertial navigation, flight control, etc. 0.5 20 25 4 Call the fault injection API to introduce errors to the primary inertial navigation, such as bias, increased noise, or complete failure 1 14 10 5 The intelligent agent analyzes the test data to determine whether the redundancy determination logic of the flight control computer is correct; 1 20 16 6 The intelligent agent calls the simulation test report template to generate a test report 0.5 40 43 Substitute the data in Table 2 into the cost function to obtain the execution cost of the i-th decision-making path as 122.4.

[0054] Normalize the cost function to obtain the optimization function of the decision path. The optimized execution cost calculated through the optimization function of the decision path is 0.312.

[0055] Through the scoring function of the decision path Calculate that the total score of the decision path generated in this simulation experiment is 3.7903.

[0056] According to the scoring judgment table, evaluate the decision path of this simulation experiment. The scoring judgment table is shown in Table 3.

[0057] Table 3: Scoring Judgment Table Average score Judgment result Strategy Below 0.4 Invalid decision chain Discard 0.4—0.6 Partially valid decision chain Update and enter the decision chain, mark it as optimizable, and continue to evaluate in subsequent test training 0.6 Valid decision chain Update and enter the decision chain It can be seen from Table 3 that the judgment result of the decision path in this simulation experiment is valid.

[0058] Use the effectiveness pruning formula to perform 200 tests on the decision path of the decision-making layer of the intelligent agent, and record the execution results as: success: 191 times, failure: 9 times. Thus, it is judged that this decision path does not meet the pruning conditions and no pruning is performed.

[0059] The present invention uses the flight control simulation intelligent agent to conduct experiment design, experiment execution control, and result analysis for the flight control semi-physical simulation test experiment environment, so as to achieve the purpose of autonomously completing the simulation test experiment without relying on human power, avoiding the human errors introduced in the simulation test process of the flight control computer, and improving the reliability of the simulation experiment results.

[0060] See Figure 2 , the present invention also provides a building system of a flight control computer simulation test system formed based on the building method of the above-mentioned flight control computer simulation test system, including a construction module, a first building module, a second building module, and a system formation module; Construction module: used to construct the structured simulation test knowledge base of the aircraft; First building module: used to build the flight control semi-physical simulation test experiment environment and the intelligent agent that operates and perceives the flight control semi-physical simulation test experiment environment, wherein the intelligent agent is trained through the structured simulation test knowledge base of the aircraft; Second building module: used to build a perception layer, a decision-making layer, a memory layer, and an execution layer for the intelligent agent to form a flight control simulation intelligent agent; wherein, the decision-making layer of the intelligent agent uses the scoring function and the optimization function of the decision path to screen the decision path, and then uses the dynamic pruning strategy to eliminate the decision paths that do not meet the dynamic pruning conditions after screening. The dynamic pruning conditions include the decision path threshold and the number of executions; the decision path refers to the flight path of the aircraft in the simulation test experiment; System formation module: used to physically connect the flight control hardware-in-the-loop simulation test environment with the flight control simulation agent to form a flight control computer simulation test system.

[0061] The construction system of the flight control computer simulation test system of the present invention corresponds to the above-mentioned construction method of the flight control computer simulation test system. Among them, for the specific content of the construction module, the first construction module, the second construction module, and the system formation module, please refer to the description in the above-mentioned part of the construction method of the flight control computer simulation test system, and the present invention will not elaborate here.

[0062] The present invention also provides a memory storing program files, and the program files are executed to implement the program instructions formed by the above-mentioned construction method of the flight control computer simulation test system. Specifically, for the construction method of the flight control computer simulation test system, please refer to the above description.

[0063] The memory in the present invention may specifically include random access memory (RAM), internal memory, read-only memory (ROM), programmable ROM, erasable programmable ROM, registers, hard disks, removable disks, or CD-ROMs. It should be noted that those skilled in the art can specifically select the form and type of the storage medium according to actual usage requirements, and the present invention does not make further specific limitations.

[0064] The present invention also provides an electronic device including a processor and a memory coupled to each other. Among them, the memory is used to store the program instructions formed by the above-mentioned construction method of the flight control computer simulation test system; the processor is used to execute the program instructions stored in the memory. Specifically, for the construction method of the flight control computer simulation test system, please refer to the above description.

[0065] The electronic device in the present invention includes any electronic device such as a computer, a mobile terminal, a remote control device, or a wearable device that can execute program instructions.

[0066] The above content is only used to illustrate the technical solution of the present invention, rather than to limit the present invention; although the present invention has been described in detail with reference to the foregoing, those of ordinary skill in the art should understand that they can still modify the foregoing described technical solution, or perform equivalent substitution on some or all of its technical features; and these modifications or substitutions do not make the essence of the corresponding technical solution deviate from the scope of the technical solution of the present invention.

Claims

1. A method for building a flight control computer simulation test system, characterized in that, Including: Constructing a structured simulation test knowledge base for the aircraft; Setting up a flight control hardware-in-the-loop simulation test environment and an agent for operating and perceiving the flight control hardware-in-the-loop simulation test environment, where the agent is trained through the structured simulation test knowledge base of the aircraft; Setting up a perception layer, a decision-making layer, a memory layer, and an execution layer for the agent to form a flight control simulation agent; among them, the decision-making layer of the agent uses the scoring function of the decision-making path and the optimization function of the decision-making path to screen the decision-making paths, and then uses the dynamic pruning strategy to eliminate the decision-making paths that do not meet the dynamic pruning conditions after screening. The dynamic pruning conditions include the decision-making path threshold and the number of executions; the decision-making path refers to the flight path of the aircraft in the simulation test; Physically connecting the flight control hardware-in-the-loop simulation test environment with the flight control simulation agent to form a flight control computer simulation test system.

2. The method for building a flight control computer simulation test system according to claim 1, wherein The specific process of screening using the scoring function of the decision-making path and the optimization function of the decision-making path includes: Determining the total score of each decision-making path according to the node actions of each decision-making path and the rewards obtained after the execution of the node actions; Constructing a cost function, normalizing the cost function to form an optimization function for the decision-making path; Evaluating the total scores of all decision-making paths according to the optimization function of the decision-making path, screening the decision-making paths, and forming the screened decision-making paths.

3. The method for building the flight control computer simulation test system according to claim 2, wherein The specific process of using the dynamic pruning strategy to eliminate the decision-making paths that do not meet the dynamic pruning conditions after screening includes: Constructing an effectiveness pruning formula for the decision-making path, dynamically evaluating the screened decision-making paths according to the effectiveness pruning formula of the decision-making path, and eliminating the decision-making paths that do not meet the dynamic pruning conditions in the screened decision-making paths according to the dynamic evaluation results.

4. The method for building a flight control computer simulation test system according to claim 3, wherein The specific process of constructing the structured simulation test knowledge base for the aircraft includes: Obtaining the simulation test knowledge base of the aircraft; Using data sorting algorithms, data cleaning strategies, data tokenization algorithms, and standardization processing algorithms to classify and organize, noise cleaning, data tokenization, and data annotation of the simulation test knowledge base of the aircraft in sequence to form a structured simulation test knowledge base for the aircraft.

5. The method for building a flight control computer simulation test system according to claim 1, wherein The method for setting up the flight control computer simulation test system further includes establishing a data dictionary to associate the perception layer of the agent with the decision-making layer of the agent, so that the decision-making layer of the agent can call the information perceived in the perception layer of the agent.

6. The method for building a flight control computer simulation test system according to claim 1, wherein The method for setting up the flight control computer simulation test system further includes configuring a standard API call interface, and the flight control hardware-in-the-loop simulation test environment and the flight control simulation agent are physically connected through the standard API call interface.

7. The method for building a flight control computer simulation test system according to claim 4, characterized in that The simulation test knowledge base of the aircraft includes the initial scene information of the flight control computer, the simulation test purpose information of the flight control computer, the simulation test method information of the flight control computer, the simulation test process information of the flight control computer, the simulation test data analysis method of the flight control computer, the simulation test result judgment standard of the flight control computer, the sensor data of the flight control computer, the simulation case analysis of the flight control computer, the simulation test outline of the flight control computer, and the simulation test report of the flight control computer.

8. The building system of the flight control computer simulation test system formed by the method for building the flight control computer simulation test system according to claim 1, characterized in that, Including: A building module: used to build a structured simulation test knowledge base for the aircraft; A first building module: used to build a flight control hardware-in-the-loop simulation test environment and an intelligent agent for operating and perceiving the flight control hardware-in-the-loop simulation test environment, where the intelligent agent is trained through the structured simulation test knowledge base of the aircraft; A second building module: used to build a perception layer, a decision-making layer, a memory layer, and an execution layer for the intelligent agent to form a flight control simulation intelligent agent; among them, the decision-making layer of the intelligent agent uses a scoring function of the decision-making path and an optimization function of the decision-making path to screen the decision-making path, and then uses a dynamic pruning strategy to eliminate the decision-making paths that do not meet the dynamic pruning conditions after screening. The dynamic pruning conditions include a decision-making path threshold and the number of executions; the decision-making path refers to the flight path of the aircraft in the simulation test; And a system formation module: used to physically connect the flight control hardware-in-the-loop simulation test environment and the flight control simulation intelligent agent to form a flight control computer simulation test system.

9. A memory, characterized in that, A program file is stored, and the program file is executed to implement program instructions formed by the method for building the flight control computer simulation test system according to any one of claims 1-7.

10. An electronic device, characterized in that, Including a processor and a memory that are coupled to each other, where The memory: used to store program instructions formed by the method for building the flight control computer simulation test system according to any one of claims 1-7; The processor: used to execute the program instructions stored in the memory.

Images