Evolutionary Intelligence for automated test generation

Speaker: Pouria Derakhshanfar (TU Delft)

One of the crucial tasks in the software development process is writing test cases that identify defects in the software. Since writing test cases for different levels of testing (e.g., unit, integration, system-level) and for various testing criteria (e.g., structural coverage, mutation score) is labor-intensive and expensive, many studies introduced automated-test generation approaches and tools that use different techniques to ease these tasks for developers. Among these approaches, evolutionary-based test generation approaches show considerable potential. For over a decade, many studies have introduced and assessed these techniques for different testing levels and criteria. These studies confirm that evolutionary-based test generation approaches are able to generate tests with high coverage and fault-detection capability in complex real-world software projects. This presentation explains some of the well-known techniques introduced in this topic and their challenges and limitations. It also presents practices that were introduced to address these challenges by combining human and computational intelligence.