As is the case with any very complex and interactive software, many video games are released with various minor or major issues that can potentially affect the user experience, cause security issues for players, or exploit the companies that deliver the products. To test their games, companies invest important resources in quality assurance personnel who usually perform the testing mostly manually. The main goal of our work is to automate various parts of the testing process that involve human users (testers) and thus to reduce costs and run more tests in less time. The secondary goal is to provide mechanisms to make test specification writing easier and more efficient. We focus on solving initial real-world problems that have emerged from several discussions with industry partners. In this paper, we present RiverGame, a tool that allows game developers to automatically test their products from different points of view: the rendered output, the sound played by the game, the animation and movement of the entities, the performance and various statistical analyses. We also address the problem of input priorities, scheduling, and directing the testing effort towards custom and dynamic directions. At the core of our methods, we use state-of-the-art artificial intelligence methods for analysis and a behavior-driven development (BDD) methodology for test specifications. Our technical solution is open-source, independent of game engine, platform, and programming language.

Combining fuzzing techniques and reinforcement learning could be an important direction in software testing. However, there is a gap in support for experimentation in this field, as there are no open-source tools to let academia and industry to perform experiments easily. The purpose of this paper is to fill this gap by introducing a new framework, named RiverFuzzRL, on top of our already mature frame-work for AI-guided fuzzing, River. We provide out-of-the-box implementations for users to choose from or customize for their test target. The work presented here is performed on testing binaries and does not require access to the source code, but it can be easily adapted to other types of software testing as well. We also discuss the challenges faced, opportunities, and factors that are important for performance, as seen in the evaluation.

The diversity of agent behaviors is an important topic for the quality of video games and virtual environments in general. Offering the most compelling experience for users with different skills is a difficult task, and usually needs important manual human effort for tuning existing code. This can get even harder when dealing with adaptive difficulty systems. Our paper's main purpose is to create a framework that can automatically create behaviors for game agents of different difficulty classes and enough diversity. In parallel with this, a second purpose is to create more automated tests for showing defects in the source code or possible logic exploits with less human effort.