MazeWorld: A Game-Based Environment developed to Assess Teaming Behaviors

A testbed has been developed to study multiple constructs of teamwork in a parameterized environment. MazeWorld is a testbed to learn and automate the collection of objective team metrics based on observable behaviors. Quantifying teamwork plays an important role in future agent cooperation. Challenges include differentiating individual skills from skills of the team to not only understand the team as an entity but each of the participating members. MazeWorld is a Unity-based cooperative game that involves three distinct roles working interdependently to achieve a goal. A series of trials were run in MazeWorld to test proposed observable measures of teaming constructs. Teams performed better as the trials progressed, and differences were found between teams. This effort serves as a proof of concept that MazeWorld can explore team behavior using automated team metrics to inform future work.