Robotics agent for automated gameplay : testing and balancing

Abstract

This thesis explores the use of robotic entities and automated artificial intelligence tools in enhancing playtesting and balancing within games, especially among game development students, independent developers, and small-scale development teams with constrained technological capabilities. Contemporary games have grown in complexity, requiring sophis-ticated levels of playtesting, but most of this testing is currently carried out by playtesting by hand, often resulting in inconsistent testing, gradual progress, and unforeseen errors within games. However, this is often due to a lack of awareness about available technologies and tools capable of solving these issues related to playtesting within games, including those used by small-scale development teams. A Mixed Methods design is used in this thesis. A Systematic Literature Review of academic and industry papers helps identify the State-of-the-Art in automation related to testing games. Along with the Literature Review, an online survey involving 1,000 participants is used to understand awareness, use, perceived bene-fits, difficulties, and intentions to use AI-assisted Automated Testing tools. Results indicate a lack of use of automation despite numerous benefits such as better Test Coverage, Quick Bug Identification, and Better Balancing, which can be attributed to a lack of awareness and complexity, among other factors, despite which the willingness to use Automated plug-and-play agents is strong. The conclusion drawn is that automated gameplay testing is both beneficial and possible if tools with a reduced level of complexity were made accessible. Suggestions include the use of hybrid scripted and artificial intelligence-based agents, open-source platforms, and the design of simple automation tools targeting new developers.