Unmanned aerial vehicle task system automatic test system based on large model technology

Abstract

The invention relates to the technical field of unmanned aerial vehicle testing, in particular to an unmanned aerial vehicle task system automatic testing system based on a large model technology. Comprising a management end, a task semantic analysis module, a large model reasoning decision center, a test case generation module, a digital twin simulation test engine, a multi-protocol adaptation interface, a multi-source data acquisition module, a data fusion module, a test evaluation module and a model optimization module. The task semantic analysis module is connected with the management end, the large model reasoning decision center is connected with the task semantic analysis module, the test case generation module is connected with the large model reasoning decision center, and the digital twin simulation test engine is connected with the test case generation module. In this way, the technical problems that in the prior art, the task semantic understanding and intelligent decision-making ability of an unmanned aerial vehicle task system test mode is poor, comprehensive and optimized test cases are difficult to generate, and defects exist in the aspects of multi-source data processing and model feedback optimization are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle testing, and particularly relates to an unmanned aerial vehicle task system automatic testing system based on large model technology. BACKGROUND

[0002] With the rapid development of unmanned aerial vehicle technology, unmanned aerial vehicles have been widely used in many fields such as military reconnaissance, logistics distribution, environmental monitoring, agricultural plant protection, etc. The complexity and diversity of unmanned aerial vehicle task systems are increasing, and their performance and reliability are directly related to the success of task execution. In order to ensure that the unmanned aerial vehicle task system runs stably and efficiently in various complex environments, it is crucial to comprehensively and accurately test the unmanned aerial vehicle task system.

[0003] However, the existing unmanned aerial vehicle task system testing method has poor task semantic understanding and intelligent decision-making ability, and it is difficult to generate comprehensive and optimized test cases, and there are deficiencies in multi-source data processing and model feedback optimization. SUMMARY

[0004] The present application relates to the technical field of unmanned aerial vehicle testing, and particularly relates to an unmanned aerial vehicle task system automatic testing system based on large model technology.

[0005] To achieve the above-mentioned purpose, the present application adopts an unmanned aerial vehicle task system automatic testing system based on large model technology, which comprises a management end, a task semantic analysis module, a large model reasoning and decision center, a test case generation module, a digital twin simulation test engine, a multi-protocol adaptation interface, a multi-source data acquisition module, a data fusion module, a test evaluation module and a model optimization module. The task semantic analysis module is connected with the management end, the large model reasoning and decision center is connected with the task semantic analysis module, the test case generation module is connected with the large model reasoning and decision center, the digital twin simulation test engine is connected with the test case generation module, the multi-protocol adaptation interface is connected with the digital twin simulation test engine and the unmanned aerial vehicle task system, the multi-source data acquisition module is connected with the unmanned aerial vehicle task system, the data fusion module is connected with the multi-source data acquisition module, the test evaluation module is connected with the data fusion module, and the model optimization module is connected with the test evaluation module and the large model reasoning and decision center. The task semantic analysis module is used for receiving natural language or structured mixed description task requirements transmitted by the management end. The large model inference decision center is configured to generate a test strategy and an executable test sequence; and the test case generation module is configured to construct a test scene and generate an optimized test case. The digital twin simulation test engine is configured to perform high-fidelity simulation testing. The multi-protocol adaptation interface is connected to the digital twin simulation test engine and the actual unmanned aerial vehicle task system, supports multiple unmanned aerial vehicle communication protocols such as MAVLink, RTK, and SDK, and realizes bidirectional data interaction between the simulation environment and the real machine system. The multi-source data acquisition module is configured to acquire state data, environmental data, and communication logs in real time; and the data fusion module is configured to process multi-source data and perform feature fusion. The test evaluation module is configured to perform performance evaluation and anomaly detection; and the model optimization module is connected to the test evaluation module and the large model inference decision center to realize incremental learning based on feedback.

[0006] The unmanned aerial vehicle task system automatic test system based on large model technology further includes a blockchain storage module connected to the test evaluation module. The blockchain storage module includes a hash fingerprint generation unit, an intelligent contract interface, and a trusted report issuing unit. The hash fingerprint generation unit performs SHA-3 calculation on the test results and calibration parameters to generate a unique fingerprint. The intelligent contract interface writes the fingerprint and timestamp into the blockchain and sets a test status bit. The trusted report issuing unit automatically generates a digital authentication report when the test is passed.

[0007] The task semantic analysis module includes a natural language processing unit, a constraint extraction unit, and a semantic vectorization unit. The natural language processing unit is configured to perform word segmentation and entity recognition on the natural language or structured mixed task description input by the user. The constraint extraction unit is configured to extract airspace restrictions, time constraints, and performance index requirements in the task. The semantic vectorization unit converts the parsed task elements into high-dimensional semantic vectors for use by the large model inference decision center.

[0008] The digital twin simulation test engine has a high-fidelity dynamics calculation mechanism and a sensor error injection mechanism. The high-fidelity dynamics calculation mechanism uses aerodynamic force-propulsion-mass coupling equations to run synchronously with the on-board software of the unmanned aerial vehicle under test. The sensor error injection mechanism injects simulation signals such as GNSS drift and IMU offset according to test requirements.

[0009] The large model reasoning decision center adopts a dynamic parameter adjustment mechanism based on reinforcement learning, automatically adjusts the temperature parameter and search depth of the large model generation test strategy according to the historical test effect, and improves the coverage rate and abnormal discovery ability of the test case.

[0010] The test case generation module includes a scene construction unit, a case generation unit and a mutation optimization unit. The scene construction unit generates a simulation test scene based on the test strategy instruction. The case generation unit automatically generates specific test cases according to scene parameters. The mutation optimization unit enhances and optimizes the test cases by injecting noise, abnormal parameters or boundary conditions.

[0011] The multi-source data acquisition module includes a sensor data acquisition unit and a communication log capture unit. The sensor data acquisition unit obtains unmanned aerial vehicle attitude, position and environmental data through onboard sensors and ground stations. The communication log capture unit records all test interaction messages.

[0012] The test evaluation module includes a trajectory difference calculator and a risk weighted scorer. The trajectory difference calculator compares simulation data with expected trajectories using dynamic time warping algorithm. The risk weighted scorer calculates risk scores according to difference degree and safety criticality level. The model optimization module initiates large model incremental learning and generates compensation test cases when risk scores exceed standards.

[0013] The unmanned aerial vehicle task system automatic test system based on the large model technology can accurately receive the task demand described in natural language or structured mixed description transmitted by the management end, greatly improves the task semantic understanding ability, the large model inference decision center can generate scientific and reasonable test strategy and executable test sequence based on the analyzed task demand, enhances the intelligent decision level, and lays a foundation for generating comprehensive and optimized test cases, and the test case generation module constructs rich and varied test scenes and generates high-quality test cases accordingly. Meanwhile, the multi-source data acquisition module can acquire state data, environment data, communication logs and other multi-source data in real time, the data fusion module effectively processes and fuses the features, and solves the multi-source data processing problem. In addition, the test evaluation module carries out performance evaluation and abnormality detection, the model optimization module realizes incremental learning based on feedback in combination with the evaluation result and the large model inference decision center, improves the model feedback optimization mechanism, and comprehensively improves the efficiency and quality of the unmanned aerial vehicle task system test. In this way, the technical problems of poor task semantic understanding and intelligent decision ability in the prior art unmanned aerial vehicle task system test mode, difficulty in generating comprehensive and optimized test cases, and deficiencies in multi-source data processing and model feedback optimization are solved. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0015] Figure 1 is the principle block diagram of the first embodiment of the present application.

[0016] Figure 2 is the principle block diagram of the second embodiment of the present application.

[0017] 101 - management terminal, 102 - task semantic analysis module, 103 - large model inference decision center, 104 - test case generation module, 105 - digital twin simulation test engine, 106 - multi-protocol adaptation interface, 107 - multi-source data acquisition module, 108 - data fusion module, 109 - test evaluation module, 110 - model optimization module, 111 - blockchain storage module, 112 - hash fingerprint generation unit, 113 - smart contract interface, 114 - trusted report issuing unit, 115 - natural language processing unit, 116 - constraint extraction unit, 117 - semantic vectorization unit, 118 - scene construction unit, 119 - use case generation unit, 120 - variation optimization unit, 121 - sensor data acquisition unit, 122 - communication log capture unit, 201 - space-time alignment module, 202 - login terminal, 203 - identity verification module, 204 - permission control module. DETAILED DESCRIPTION

[0018] Embodiments of the present application will be described in detail below, examples of which are shown in the accompanying drawings, the embodiments described below by referring to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0019] First embodiment, please refer to Figure 1 , Figure 1 is the principle block diagram of the first embodiment of the present application.

[0020] The present application provides a kind of unmanned aerial vehicle task system automatic test system based on large model technology, including management terminal 101, task semantic analysis module 102, large model inference decision center 103, test case generation module 104, digital twin simulation test engine 105, multi-protocol adaptation interface 106, multi-source data acquisition module 107, data fusion module 108, test evaluation module 109, model optimization module 110 and blockchain storage module 111;The blockchain storage module 111 includes hash fingerprint generation unit 112, smart contract interface 113 and trusted report issuing unit 114;The task semantic analysis module 102 includes natural language processing unit 115, constraint extraction unit 116 and semantic vectorization unit 117;The test case generation module 104 includes scene construction unit 118, use case generation unit 119 and variation optimization unit 120;The multi-source data acquisition module 107 includes sensor data acquisition unit 121 and communication log capture unit 122; The foregoing technical solution solves the technical problems that the task semantic understanding and intelligent decision-making capability of the unmanned aerial vehicle task system test mode in the prior art are poor, it is difficult to generate comprehensive and optimized test cases, and there are deficiencies in multi-source data processing and model feedback optimization.

[0021] In this specific embodiment, the task semantic parsing module 102 is used to receive task requirements in natural language or structured mixed description from the management terminal 101. The large model inference and decision-making center 103 is used to generate test strategies and executable test sequences; the test case generation module 104 is used to construct test scenarios and generate optimized test cases. The digital twin simulation test engine 105 is used to perform high-fidelity simulation tests; The multi-protocol adapter interface 106 is connected to the digital twin simulation test engine 105 and the actual UAV mission system respectively, supporting multiple UAV communication protocols such as MAVLink, RTK and SDK, and realizing bidirectional data interaction between the simulation environment and the real system. The multi-source data acquisition module 107 is used to acquire status data, environmental data, and communication logs in real time; the data fusion module 108 is used to process multi-source data and perform feature fusion. The test evaluation module 109 is used for performance evaluation and anomaly detection; the model optimization module 110 is connected to the test evaluation module 109 and the large model inference decision center 103 to realize feedback-based incremental learning.

[0022] The large model inference decision center 103 adopts a dynamic parameter tuning mechanism based on reinforcement learning, which automatically adjusts the temperature parameters and search depth of the large model generation test strategy according to the historical test results, so as to improve the coverage of test cases and the ability to detect anomalies.

[0023] Specifically, the task semantic parsing module 102 is connected to the management terminal 101; the large model inference and decision-making center 103 is connected to the task semantic parsing module 102; the test case generation module 104 is connected to the large model inference and decision-making center 103; the digital twin simulation test engine 105 is connected to the test case generation module 104; the multi-protocol adaptation interface 106 is connected to both the digital twin simulation test engine 105 and the UAV mission system; the multi-source data acquisition module 107 is connected to the UAV mission system; the data fusion module 108 is connected to the multi-source data acquisition module 107; the test evaluation module 109 is connected to the data fusion module 108; and the model optimization module 110 is connected to both the test evaluation module 109 and the large model inference and decision-making center 103. In practical use, the task semantic parsing module 102 can accurately receive task requirements described in natural language or a mixed structure from the management terminal 101, greatly improving the task semantic understanding capability. Based on the parsed task requirements, the large model inference and decision-making center 103 can generate scientific and reasonable test strategies and executable test sequences, enhancing the level of intelligent decision-making and laying the foundation for generating comprehensive and optimized test cases. The test case generation module 104 then constructs a variety of test scenarios and generates high-quality test cases accordingly. Simultaneously, the multi-source data acquisition module 107 can collect multi-source data such as status data, environmental data, and communication logs in real time. The data fusion module 108 effectively processes and fuses these data, solving the problem of multi-source data processing. Furthermore, the test evaluation module 109 performs performance evaluation and anomaly detection, and the model optimization module 110 combines the evaluation results with the large model inference and decision-making center 103 to achieve feedback-based incremental learning, improving the model feedback optimization mechanism and comprehensively enhancing the efficiency and quality of UAV mission system testing. This approach addresses the technical issues in existing UAV mission system testing methods, such as poor task semantic understanding and intelligent decision-making capabilities, difficulty in generating comprehensive and optimized test cases, and deficiencies in multi-source data processing and model feedback optimization.

[0024] Secondly, the hash fingerprint generation unit 112 performs SHA-3 calculation on the test results and calibration parameters to generate a unique fingerprint; The smart contract interface 113 writes the fingerprint and timestamp into the blockchain and sets the test status bit; The trusted report issuing unit 114 automatically generates a digital authentication report when the test is passed.

[0025] Meanwhile, the natural language processing unit 115 is used to perform word segmentation and entity recognition on the natural language or structured hybrid task description input by the user; The constraint extraction unit 116 is used to extract the spatial restrictions, time constraints and performance requirements in the task. The semantic vectorization unit 117 converts the parsed task elements into high-dimensional semantic vectors for use by the large model reasoning and decision center 103.

[0026] Furthermore, the high-fidelity dynamics calculation mechanism employs aerodynamic-propulsion-mass coupling equations and operates synchronously with the onboard software of the UAV under test; The sensor error injection mechanism injects simulated signals such as GNSS drift and IMU offset according to test requirements.

[0027] Meanwhile, the scenario construction unit 118 generates a simulation test scenario based on the test strategy instructions; The test case generation unit 119 automatically generates specific test cases based on scenario parameters; The mutation optimization unit 120 enhances and optimizes test cases by injecting noise, abnormal parameters, or boundary conditions.

[0028] The sensor data acquisition unit 121 acquires UAV attitude, position and environmental data through airborne sensors and ground station; The communication log capture unit 122 records all test interaction messages.

[0029] Furthermore, the test evaluation module 109 includes a trajectory difference calculator and a risk-weighted scorer; The trajectory difference calculator uses a dynamic time warping algorithm to compare simulation data with the expected trajectory. The risk weighted scorer calculates a risk score based on the degree of difference and the level of safety criticality. When the risk score exceeds the limit, the model optimization module 110 initiates large model incremental learning and generates compensation test cases.

[0030] In this embodiment of the automated testing system for UAV mission systems based on large model technology, the task semantic parsing module 102 accurately receives task requirements from the management terminal 101 in natural language or a mixed structured description, greatly improving task semantic understanding capabilities. Based on the parsed task requirements, the large model inference and decision-making center 103 generates scientifically sound test strategies and executable test sequences, enhancing intelligent decision-making and laying the foundation for generating comprehensive and optimized test cases. The test case generation module 104 then constructs diverse test scenarios and generates high-quality test cases. Simultaneously, the multi-source data acquisition module 107 collects multi-source data such as status data, environmental data, and communication logs in real time. The data fusion module 108 effectively processes and fuses these data, solving the problem of multi-source data processing. Furthermore, the test evaluation module 109 performs performance evaluation and anomaly detection, and the model optimization module 110 combines the evaluation results with the large model inference and decision-making center 103 to achieve feedback-based incremental learning, improving the model feedback optimization mechanism and comprehensively enhancing the efficiency and quality of UAV mission system testing. This approach addresses the technical issues in existing UAV mission system testing methods, such as poor task semantic understanding and intelligent decision-making capabilities, difficulty in generating comprehensive and optimized test cases, and deficiencies in multi-source data processing and model feedback optimization.

[0031] Second embodiment, please refer to Figure 2 , Figure 2 This is a schematic diagram of the second embodiment of the present invention.

[0032] In addition to the technology of the first embodiment, the present invention provides an automatic testing system for unmanned aerial vehicle mission systems based on large model technology, which also includes a spatiotemporal alignment module 201, a login terminal 202, an authentication module 203, and an access control module 204.

[0033] In this specific implementation, the spatiotemporal alignment module 201 is connected to the data fusion module 108. The spatiotemporal alignment module 201 adopts a unified spatiotemporal reference algorithm based on extrinsic parameter calibration and PTP time synchronization, combined with the ICP point cloud registration algorithm, to eliminate spatial deviations. Then, the PTP protocol is used to achieve microsecond-level time synchronization to ensure strict alignment of timestamps of different sensor data. Finally, the ICP algorithm is used to iteratively register point cloud data in dynamic scenes to compensate for dynamic errors caused by UAV movement.

[0034] The login terminal 202 is connected to the management terminal 101, the authentication module 203 is connected to the login terminal 202, and the access control module 204 is connected to the authentication module 203. The login terminal 202 transmits user credentials through a TLS encrypted channel. The authentication module 203 performs two-factor authentication by combining static passwords with dynamic tokens (such as TOTP time synchronization tokens) to effectively resist brute-force attacks and man-in-the-middle attacks. The access control module 204 is based on the RBAC model and dynamically allocates system operation permissions according to user roles (such as administrators, testers, and observers), and prevents permission abuse through permission inheritance and constraint mechanisms.

[0035] This embodiment of an automated testing system for UAV mission systems based on large model technology employs a spatiotemporal alignment module 201 that uses a unified spatiotemporal reference algorithm based on extrinsic parameter calibration and PTP time synchronization, combined with an ICP point cloud registration algorithm to eliminate spatial deviations. It then utilizes the PTP protocol to achieve microsecond-level time synchronization, ensuring strict alignment of timestamps from different sensors. Finally, it iteratively registers point cloud data in dynamic scenes using the ICP algorithm to compensate for dynamic errors caused by UAV movement. The login terminal 202 transmits user credentials via a TLS encrypted channel. The authentication module 203 combines static passwords with dynamic tokens (such as TOTP time synchronization tokens) for two-factor authentication, effectively resisting brute-force attacks and man-in-the-middle attacks. The access control module 204, based on the RBAC model, dynamically allocates system operation permissions according to user roles (such as administrator, tester, and observer), and prevents permission abuse through permission inheritance and constraint mechanisms.

[0036] The above description discloses only one preferred embodiment of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. An automated testing system for unmanned aerial vehicle (UAV) mission systems based on large-scale model technology, characterized in that, It includes a management terminal, a task semantic parsing module, a large model inference and decision-making center, a test case generation module, a digital twin simulation test engine, a multi-protocol adaptation interface, a multi-source data acquisition module, a data fusion module, a test evaluation module, and a model optimization module; The task semantic parsing module is connected to the management terminal; the large model inference and decision-making center is connected to the task semantic parsing module; the test case generation module is connected to the large model inference and decision-making center; the digital twin simulation test engine is connected to the test case generation module; the multi-protocol adaptation interface is connected to both the digital twin simulation test engine and the UAV mission system; the multi-source data acquisition module is connected to the UAV mission system; the data fusion module is connected to the multi-source data acquisition module; the test evaluation module is connected to the data fusion module; and the model optimization module is connected to both the test evaluation module and the large model inference and decision-making center. The task semantic parsing module is used to receive task requirements described in natural language or structured mixed description from the management terminal; The large model inference and decision-making center is used to generate test strategies and executable test sequences; the test case generation module is used to construct test scenarios and generate optimized test cases. The digital twin simulation test engine is used to perform high-fidelity simulation tests; The multi-protocol adaptation interface is connected to the digital twin simulation test engine and the actual UAV mission system respectively, supporting multiple UAV communication protocols such as MAVLink, RTK and SDK, to realize bidirectional data interaction between the simulation environment and the real system. The multi-source data acquisition module is used to collect status data, environmental data, and communication logs in real time; the data fusion module is used to process multi-source data and perform feature fusion. The test evaluation module is used for performance evaluation and anomaly detection; the model optimization module is connected to the test evaluation module and the large model inference decision center to realize incremental learning based on feedback.

2. The automated testing system for UAV mission systems based on large-scale model technology as described in claim 1, characterized in that, The automated testing system for unmanned aerial vehicle mission systems based on large model technology also includes a blockchain storage module, which is connected to the testing and evaluation module. The blockchain storage module includes a hash fingerprint generation unit, a smart contract interface, and a trusted report issuance unit. The hash fingerprint generation unit performs SHA-3 calculation on the test results and calibration parameters to generate a unique fingerprint; The smart contract interface writes the fingerprint and timestamp into the blockchain and sets the test status bit. The trusted report issuing unit automatically generates a digital authentication report when the test is passed.

3. The automated testing system for UAV mission systems based on large-scale model technology as described in claim 2, characterized in that, The task semantic parsing module includes a natural language processing unit, a constraint extraction unit, and a semantic vectorization unit; The natural language processing unit is used to perform word segmentation and entity recognition on the natural language or structured hybrid task description input by the user; The constraint extraction unit is used to extract spatial restrictions, time constraints, and performance requirements in the task. The semantic vectorization unit transforms the parsed task elements into high-dimensional semantic vectors for use by the large model inference and decision-making center.

4. The automated testing system for UAV mission systems based on large-scale model technology as described in claim 3, characterized in that, The digital twin simulation test engine features a high-fidelity dynamics calculation mechanism and a sensor error injection mechanism. The high-fidelity dynamics calculation mechanism employs aerodynamic-propulsion-mass coupling equations and runs synchronously with the onboard software of the UAV under test. The sensor error injection mechanism injects simulated signals such as GNSS drift and IMU offset according to test requirements.

5. The automated testing system for UAV mission systems based on large-scale model technology as described in claim 4, characterized in that, The large model inference decision center adopts a dynamic parameter tuning mechanism based on reinforcement learning. It automatically adjusts the temperature parameters and search depth of the large model's test generation strategy according to historical test results to improve test case coverage and anomaly detection capabilities.

6. The automated testing system for UAV mission systems based on large-scale model technology as described in claim 5, characterized in that, The test case generation module includes a scenario construction unit, a test case generation unit, and a mutation optimization unit; The scenario construction unit generates simulation test scenarios based on test strategy instructions; The test case generation unit automatically generates specific test cases based on scenario parameters; The mutation optimization unit enhances and optimizes test cases by injecting noise, abnormal parameters, or boundary conditions.

7. The automated testing system for UAV mission systems based on large-scale model technology as described in claim 6, characterized in that, The multi-source data acquisition module includes a sensor data acquisition unit and a communication log capture unit; The sensor data acquisition unit acquires UAV attitude, position, and environmental data through airborne sensors and ground stations; The communication log capture unit records all test interaction messages.

8. The automated testing system for unmanned aerial vehicle (UAV) mission systems based on large-scale model technology as described in claim 7, characterized in that, The test evaluation module includes a trajectory difference calculator and a risk weighted scorer; The trajectory difference calculator uses a dynamic time warping algorithm to compare simulation data with the expected trajectory. The risk weighted scorer calculates a risk score based on the degree of difference and the level of safety criticality. When the risk score exceeds the limit, the model optimization module initiates large-scale incremental learning of the model and generates compensation test cases.

Images