Automated testing of computer games presents unique challenges due to their complexity, dynamic nature, and pervasive randomness, making traditional static tests (e.g., sequences of events) fragile and inadequate for thorough exploration or bug detection. While Reinforcement Learning has shown promise in game playing, its agents are usually trained on game-specific objectives (winning) over systematic bug exploration, and demand extensive training and hand-crafted reward functions that may struggle with frequent game changes during development. This talk argues against abandoning traditional automated test generation in the context of game testing. Instead, we propose a hybrid approach that combines white-box testing, specifically code coverage-guided test generation, with reinforcement learning in terms of neuroevolution. This methodology generates dynamic test suites—machine learning models that learn to play games implicitly, driven by the goal to explore their source code and to maximize code coverage.
Chair, Faculty of Computer Science and Mathematics
University of Passau, Germany
Chair of Software Engineering II
Gordon Fraser is a full professor in Computer Science at the University of Passau, Germany. He received a PhD in computer science from Graz University of Technology, Austria, in 2007, worked as a post-doc at Saarland University, and was a Senior Lecturer at the University of Sheffield, UK. The central theme of his research is improving software quality, and his recent research concerns the prevention, detection, and removal of defects in software. He is known for his work on search-based testing, in which he created popular unit test generators such as EvoSuite or Pynguin.
