DSLs for game development and AI for video game testing

by [Emmanuel Chebi Hugo Martin]
28/11/2019
DiverSE Coffee
Rennes, France

Abstract

**Emmanuel** : Developing a game requires a lot of different skills and thus involves many people specialized in different domains such as game design, AI or storytelling and that may not be used to develop software. As a consequence, game development should be a breeding ground for the creation of new DSLs. We'll discuss different use cases and challenges related to DSL in the game industry, and finally why there is so few good open source games. **Hugo** :Testing video games, as of today, heavily relies on humans, the testing process being at best semi-automated. Wuji is a tool relying on Deep Reinforcement Learning to automate the testing for some parts of the video games, using multi-objective optimization to progress in the game as well as exploring the space of possibilities available in the game. Using this tool on commercial games allowed discovering previously unknown bugs.