Improving Requirement Traceability by Leveraging Video Game Simulations in Search-Based Software Engineering

Video games pose different challenges during development and maintenance than classic software. For example, common and widespread assets, that are typically created as part of video game development are Non-Player Characters (NPCs). NPCs contribute to different aspects such as storytelling and user experience, and they are typically controlled by the CPU. We theorize that a reproduction of the actions of NPCs within the game (i.e., simulations) holds key information for Game Software Engineering (GSE) tasks such as Traceability Link Recovery (TLR). This paper presents our approach for supporting TLR in GSE by leveraging video game simulations. Simulation data from NPCs is used to reduce the search space. Since the reduced search space might still be too large for manual inspection, an evolutionary TLR procedure evolves a population of code fragments. As a result, a ranking of code fragments that map the requirement to the code is obtained. We evaluate our approach in Kromaia, a commercial video game released on PC and PlayStation 4. We compare our approach against a baseline that does not incorporate simulations by means of a statistical analysis. Our approach reduces the search space by 99.21% on average, and significantly outperforms the baseline with large differences in all performance indicators. A focus group with professional developers has confirmed the acceptance of our approach. Our work provides a new direction in TLR, which is an essential task in not only GSE but also in classic software engineering.