AstroBug: Automatic Game Bug Detection Using Deep Learning

Abstract

Traditional methods of video game bug detection, such as manual testing, have been effective, but they can also be time-consuming and costly. While automated bug detection techniques hold great promise for improving testing, they still face several challenges that need to be addressed to be effective in practice. In this work, we introduce a new framework to detect perceptual bugs using a long short-term memory network, which detects bugs in games as anomalies. The detected buggy frames are then clustered to determine the category of the occurred bug. The framework was evaluated on two first person shooter games. We further enhanced the framework by implementing a reinforcement learning agent to autonomously gather datasets, effectively addressing the need for human players to collect data and manually browse through games. The enhancement was performed on a role-playing game. The outcomes obtained validate the effectiveness of the framework.