Background: Automated testing is widely used in modern software development to check if the software, including its graphical user interface (GUI), meets the expectations in terms of quality and functionality. GUI-based test automation, like other automation, aims to save time and money compared to manual testing without reducing the software quality. While automation has successfully reduced costs for other types of testing (e.g., unit-or integration tests), GUI-based testing has faced technical challenges, some of which have lingered for over a decade. 

Objective: This thesis work aims to contribute to the software engineering body of knowledge by (1) identifying the main challenges in GUI-based test automation and (2) finding technical solutions to mitigate some of the main challenges. One such challenge is to reliably identify GUI elements during test execution to prevent unnecessary repairs. Another problem is the demand for test automation and programming skills when designing stable automated tests at scale. 

Method: We conducted several studies by adopting a multi-methodological approach. First, we performed a systematic literature review to identify the main challenges in GUI-based test automation, followed by multiple studies that propose and evaluate novel approaches to mitigate the main challenges. 

Results: Our first contribution is mapping the challenges in GUI-based test automation reported in academic literature. We mapped the main challenges (i.e. most reported) on a timeline and classified them as essential or accidental. This classification is valuable since future research can focus on the main challenges that we are more likely to mitigate using a technical solution (i.e., accidental). Our second contribution is several approaches that explore novel concepts or advance state-of-the-art techniques to mitigate some of the main accidental challenges. Testing an application through an augmented layer (Augmented Testing) can reduce the demand for test automation and programming skills and mitigate the challenges of creating and maintaining model based tests. Our proposed approach for locating web elements (Similo) can increase the robustness of automated test execution. 

Conclusion: Our results provide alternative approaches and concepts that can mitigate some of the main accidental challenges in GUI-based test automation. With a more robust test execution and tool support for test modeling, we can help reduce the manual labor spent on creating and maintaining automated GUI-based tests. With a reduced cost of automation, testers can focus more on other tasks like requirements, test design, and exploratory testing.