부산대학교 도서관

Automating software testing looks forward to speeding up testing processes and ensuring possible replication of discovered software bugs. However, Automating the GUI testing process is highly challenging due to the need for human intervention to determine actions and assess outcomes. We introduce a novel approach to fully automate GUI testing using deep reinforcement learning. Our deep reinforcement learning model discovers all system states and determines possible testing sequences. The automated testing agent starts with exploring the tested environment to learn the most efficient paths for reaching maximum coverage while discovering GUI bugs. In this case, testers could focus more on functionality testing to improve the overall software quality. We evaluated the developed model on a couple of industry products, and it showed a substantial increase in coverage than random testing.