Evaluation of Automatic Test Case Generation for Android Operating System using Deep Reinforcement Learning

The industry of large-scale software for mobile devices, such as the Android operating system, presents significant challenges regarding software validation and testing. This is due to the need to test on various devices, operating system versions, connections, and different hardware configurations. As a result, the manual creation of test cases can be a time-consuming process, and test cases can become outdated with updates in the Android version. To tackle these challenges, automatic test case generation emerges as an effective solution to streamline test creation and updates. In this context, Artificial Intelligence (AI) techniques, such as Deep Reinforcement Learning (DRL), have been explored to optimize this process and ensure adequate coverage of system requirements. This study evaluated the performance of the DRL state-of-the-art tool for test case generation DRL-MOBTEST [3] in an industry scenario context to generate test cases for Android functional applications (apps). The tool was performed in nine native apps (clock, maps, calculator, wallpaper, calendar, contacts, YouTube, drive, and files) regarding the functionalities coverage. The results showed a coverage range of 74.43%, and we compared it with the random Android SDK tool Monkey in five applications, revealing a trend of 63.52% improvement. The DRL-MOBTEST tool achieved the coverage of basic application paths through the creation of different test input types, such as symbols, numbers, and letters. It enables professionals to focus on complex scenarios and improve software quality across different devices and hardware configurations. However, it's worth noting that human supervision is still necessary despite the advances offered by automated tools.