A Behavior-Driven Development and Reinforcement Learning Approach for Videogame Automated Testing

A Behavior-Driven Development and Reinforcement Learning Approach for Videogame Automated Testing

연구 분야: Software Development

학회: 2024 IEEE/ACM 8th International Workshop on Games and Software Engineering (GAS)

Video game development has undergone a significant transformation in the last decade, with modern games becoming increasingly complex and sophisticated. Testing these games for quality assurance is challenging and time-consuming, often relying on manual testers. In this paper, we introduce an automated testing approach that combines Behavior-Driven Development (BDD) with Reinforcement Learning (RL) to streamline the testing process. We present a framework that uses natural language-based test cases to describe game behaviors and expected outcomes, combined with RL, to test games automatically. We validated our approach through tests on four distinct Python-based games. We analyzed the impact of game complexity on training duration and discussed the challenges of defining optimal reward functions. Our framework provides a structured approach to address RL complexities, simplifying the process of creating test scenarios. Combining BDD and RL offers a promising solution to test complex modern video games more efficiently and ensure higher game quality upon release.