CN-121979706-A - Thermal control unmanned test system and method based on large model game and RPA

CN121979706ACN 121979706 ACN121979706 ACN 121979706ACN-121979706-A

Abstract

A thermal control unmanned test system and method based on large model game and RPA belongs to the technical field of thermal control test, and comprises the steps of analyzing a test flow prompt word, generating a next test instruction in combination with a current test state, calling a corresponding bottom test function, generating an execution instruction containing operation semantics, limiting a test boundary, inquiring and merging test data in a database to form state feedback containing a test result, intercepting a screen image according to the execution instruction, positioning position coordinates and color characteristics of a target control, executing simulated user operation on a front end interface according to the position coordinates and color information of the target control, storing a semantic interaction record which is carried out between a flow driving module and a flow execution module in a natural language form, and starting, monitoring and stopping a test flow according to an image identification positioning result or an execution state scheduling, wherein the semantic interaction record comprises the state feedback. The invention solves the problem that the current thermal control automation software can only complete the automatic execution of part of the flow.

Inventors

QU XIAOYU
WANG WEIQI
LI YONG
ZHANG WEI
YOU JIA
XU KAIHANG
DENG YUE
HE LINLU
LI TIANNAN

Assignees

中国空间技术研究院

Dates

Publication Date: 20260505
Application Date: 20251211

Claims (10)

1. Thermal control unmanned test system based on big model recreation and RPA, its characterized in that includes: The flow driving module is used for analyzing the test flow prompt word, generating a next test instruction by combining the current test state and sending the next test instruction to the flow execution module; the flow execution module is semantically isolated from the flow driving module, receives the test instruction, calls a corresponding bottom layer test function, generates an execution instruction containing operation semantics, sends the execution instruction to the RPA operation module, limits a test boundary, and inquires and merges test data in a database to form state feedback containing a test result; The RPA operation module is connected with the flow execution module and is used for executing the simulation user operation on the front-end interface according to the execution instruction and the position coordinates and the color information fed back by the image recognition positioning module; The image recognition positioning module is connected with the RPA operation module and is used for recognizing the screen image intercepted by the RPA operation module, positioning the position coordinates and the color characteristics of the target control and feeding back the position coordinates and the color information to the RPA operation module; The scheduling module is connected with the flow driving module and the flow executing module and is used for storing semantic interaction records between the flow driving module and the flow executing module in a natural language form and scheduling the starting, monitoring and ending of a test flow according to the identification result of the image identification and positioning module or the execution state of the flow executing module, wherein the semantic interaction records comprise the state feedback.
2. The thermal control unmanned test system based on the large model game and the RPA, which is disclosed in claim 1, is characterized in that the flow driving module and the flow executing module are respectively realized by a driving agent and an executing agent which are constructed based on the large model, wherein the driving agent is used for analyzing and understanding the prompt word of the test flow and generating a test instruction, the executing agent is used for analyzing the test instruction and controlling the RPA operation, and the driving agent and the executing agent interact through natural language text.
3. The thermal unmanned testing system based on large model gaming and RPA of claim 1, wherein the image recognition and positioning module is based on OpenCV implementation for performing target detection and feature extraction on centered buttons, dialog boxes, progress bars, and feature icons.
4. The large model game and RPA based unmanned thermal control test system of claim 1, wherein the test procedure comprises a functional test, logic verification, or parameter tuning procedure of a aerospace thermal control system loop.
5. The thermal unmanned testing system based on large model gaming and RPA of claim 1, wherein the user operations include capturing screen images, clicking buttons, inputting text, reading data, performing adaptive operations according to position coordinates fed back by the image recognition positioning module, and monitoring the testing progress according to color characteristics fed back by the image recognition positioning module.
6. The large model gaming and RPA based thermal unmanned test system of claim 1, wherein the RPA operating module operates in a non-intrusive manner at an operating system layer or an application layer.
7. A thermal unmanned test method implemented by the thermal unmanned test system based on large model gaming and RPA according to any one of claims 1 to 6, comprising: Analyzing the prompt word of the test flow, and generating a next test instruction by combining the current test state; according to the test instruction, a corresponding bottom layer test function is called, an execution instruction containing operation semantics is generated, a test boundary is limited, and test data in a database are queried and combined to form state feedback containing a test result; intercepting a screen image according to the execution instruction, positioning the position coordinates and color characteristics of the target control, and executing simulation user operation on a front-end interface according to the position coordinates and color information of the target control; And storing semantic interaction records between the flow driving module and the flow executing module in a natural language form, and scheduling the starting, monitoring and ending of the test flow according to the image recognition positioning result or the execution state, wherein the semantic interaction records comprise the state feedback.
8. The thermal unmanned testing method based on large model games and RPA according to claim 7, wherein the position coordinates and color features of the positioning target control are implemented based on OpenCV for target detection and feature extraction of centered buttons, dialog boxes, progress bars and feature icons.
9. The method for unmanned thermal control testing based on large model gaming and RPA of claim 7, wherein the testing procedure comprises a functional testing, logic verification or parameter tuning procedure of a aerospace thermal control system loop.
10. The large model game and RPA based thermal unmanned test method of claim 7, wherein the user operation comprises capturing screen images, clicking buttons, inputting text, reading data, performing an adaptive operation according to position coordinates, and monitoring the test progress according to color characteristics.

Description

Thermal control unmanned test system and method based on large model game and RPA Technical Field The invention relates to a thermal control unmanned test system and a thermal control unmanned test method based on a large model game and RPA, and belongs to the technical field of thermal control tests. Background At present, in the aspect of thermal control testing of a spacecraft, a software testing method based on structural data and temperature inverse solution is preliminarily formed, automatic testing is preliminarily realized through basic software, and automatic sending instruction interpretation telemetry is realized through a remote control telemetry interface with a current testing system. However, the current running thermal control automatic test system has low intelligent level and faces the following difficulties: 1) Manually screening a test loop and editing structured data as input; 2) The test flow is started and terminated manually, and the test procedure is controlled by the testers, although the problem that a large number of instructions are not required to be sent manually is solved. It can be seen that the existing thermal control automation software can only complete the automatic execution of part of the processes, cannot meet the connection requirement between the processes, and needs to manually perform function combination connection. Disclosure of Invention The invention solves the technical problems of overcoming the defects of the prior art, providing a thermal control unmanned test system and a thermal control unmanned test method based on large model game and RPA, and solving the problem that the current thermal control automatic software can only complete the automatic execution of part of the flow. The technical scheme of the invention is that in the first aspect, the thermal control unmanned test system based on the large model game and the RPA comprises: The flow driving module is used for analyzing the test flow prompt word, generating a next test instruction by combining the current test state and sending the next test instruction to the flow execution module; the flow execution module is semantically isolated from the flow driving module, receives the test instruction, calls a corresponding bottom layer test function, generates an execution instruction containing operation semantics, sends the execution instruction to the RPA operation module, limits a test boundary, and inquires and merges test data in a database to form state feedback containing a test result; The RPA operation module is connected with the flow execution module and is used for executing the simulation user operation on the front-end interface according to the execution instruction and the position coordinates and the color information fed back by the image recognition positioning module; The image recognition positioning module is connected with the RPA operation module and is used for recognizing the screen image intercepted by the RPA operation module, positioning the position coordinates and the color characteristics of the target control and feeding back the position coordinates and the color information to the RPA operation module; The scheduling module is connected with the flow driving module and the flow executing module and is used for storing semantic interaction records between the flow driving module and the flow executing module in a natural language form and scheduling the starting, monitoring and ending of a test flow according to the identification result of the image identification and positioning module or the execution state of the flow executing module, wherein the semantic interaction records comprise the state feedback. The flow driving module and the flow executing module are respectively realized by a driving agent and an executing agent which are constructed based on a large model, the driving agent is used for analyzing and understanding the prompt word of the test flow and generating a test instruction, the executing agent is used for analyzing the test instruction and controlling RPA operation, and the driving agent and the executing agent interact through natural language texts. Further, the image recognition positioning module is based on OpenCV realization and is used for carrying out target detection and feature extraction on centered buttons, dialog boxes, progress bars and feature icons. Further, the test flow comprises a function test, logic verification or parameter setting flow of the aerospace thermal control system loop. Further, the user operation comprises the steps of intercepting screen images, clicking buttons, inputting texts and reading data, the self-adaptive operation is executed according to the position coordinates fed back by the image recognition positioning module, and the test progress is monitored according to the color characteristics fed back by the image recognition positioning module. Further, the RPA operation module runs in a non-invasive m