Design And Implementation Of Game AI Effect Feedback System Based On Labeling And Visualization

With the development of artificial intelligence(AI)technology,people apply artificial intelligence technology to industries such as finance,medical care,education,and security,including the gaming industry.There are certain limitations in judging the quality of artificial intelligence algorithms by traditional evaluation indicators such as correct rate and recall rate.For example,in Multiplayer Online Battle Arena(MOBA)games,AI R&D personnel cannot intuitively understand the capabilities of game AI.In response to this problem,a game AI effect feedback system was designed to help AI R&D personnel evaluate AI capabilities through engineering methods,identify existing problems,determine direction of model improvement,and ultimately achieve continuous improvement of game AI capabilities.It can also be used as a data base for game planner to commercialize game AI.In this paper,a game AI effect feedback system based on labeling and visualization is proposed.The data annotation method is used to evaluate the rationality of AI behavior in specific scenes.The data analysis and visualization methods are used to demonstrate AI capabilities in all aspects.The system includes five functional modules: battle management,video and log transferring,annotation management,data analysis and visualization,user and authority management.The battle management module applies the AI to the game to play the battle,where the battle data is obtained.This module includes functions such as creating battles,run a match,monitoring and alerting,etc.The video and log transferring module converts the game result into a playable video file and structured battle data.The structured battle data is used for subsequent analysis.This module includes functions such as creating video transferring tasks,video transferring task schedules,etc.The annotation management module can label AI behaviors in the game,including creating label tasks,data labeling,and labeling result downloads,etc.The data analysis and visualization module parses game data into the database and displays it visually.The user and authority management module helps the administrator to manage the system users and do authority control.In this paper,the requirements analysis,overall design,detailed design and implementation of the game AI effect feedback system are introduced in detail,using the unified modeling language(UML)to elaborate.The system Web framework uses a simple and efficient PHP framework,Code Igniter.The front end uses Bootstrap,JQuery and other technologies to improve the user experience and reduce development costs.The system script part is mainly implemented using Python and Shell languages.Finally,the system was tested for its functions and performance.The integrity and correctness of the system's functionality have been determined,and the performance meets the requirements,satisfies the user's needs,and has met the requirements for stable on-line operation.