Improving the automated testing of computer games

电脑游戏是一种快速发展的应用类型。预计到2020年，游戏行业的收入将接近1600亿美元，远远超过其他流行的娱乐行业，如电影行业。由于游戏行业的规模，以及玩家很难取悦的事实，开发一款成功的游戏变得具有挑战性，而且往往成本高昂。

Computer games are a rapidly growing application genre. With a projected revenue of almost $160 billion in 2020, the gaming industry has significantly outgrown other popular entertainment industries, such as the movie industry. Because of the scale of the gaming industry, and the fact that gamers are extremely hard to please, it has become challenging, and often costly, to develop a successful game. Because of the pickiness of gamers, and the intense competition between game developers, it is essential to exhaustively test a game before its release. However, as modern games tend to be complex and have many possible in-game event sequences (i.e., interactions with the main character(s)), creating an exhaustive test suite for a game is not trivial and time consuming. Ideally, the process of creating and executing game test suites can be automated. Unfortunately, automated testing of games is extremely difficult and not often done in practice. In this project, we propose to improve the state-of-the-art of automated testing of computer games by leveraging data to help better focus the automated test generation. Our goals are to (1) understand current industrial game testing practices, (2) adapt approaches for automated gameplay to support automated test generation, (3) prioritize generated test cases that need test oracles (i.e., expected test outcomes) and (4) prioritize failing test cases that need to be dealt with by the quality assurance team.