Q-funcT: A Reinforcement Learning Approach for Automated Black Box Functionality Testing
Yadini Perez Lopez, Juan Gabriel Colonna, Edson de Araújo Silva, Richard Hada Degaki, Javier Martinez Silva · 2022
The steady growth of mobile applications users over the past years has resulted in increasing the workload of Software Quality Assurance teams. Regarding this, automated Android Testing has become a highlighted research subject. However, the state-of-art academic and industrial solutions available have mainly focused in exploratory or Automated Input Generation approaches, fewer works have addressed the challenge of automated functionality testing. Moreover, the proposed solutions exhibit several limitations standing out the vulnerability to app evolution and fragmentation. In this work we propose Q-funcT, a Reinforcement Learning based approach that aims to improve automated functionality testing by increasing portability. When compared with Scripted Test Cases our method takes a few minutes longer to complete the defined missions; however, regarding portability, Q-funcT shown a notably better performance.