Role-Based Multiplayer Content Online Adaptation in Large-Scale Scenarios.

Many multiplayer games offer their players different distinct roles to choose from. Especially cooperative games use roles, as these foster cooperation between the players. Due to the complexity of possible group compositions their choice however, is often limited, as specific key roles are required based on the presented challenge. This limitation can severely reduce player enjoyment if it prevents them from playing their favourite role. Manually adapting every challenge for every possible group composition is infeasible due to the high amount of group compositions and their role interactions. These aspects are particularly relevant in the context of game-based learning, when adapting content to the current group of players. To offer a feasible solution to the problem, this paper examines how player roles and interactions can be formalized and used in an automatic adaptation algorithm. Here, role properties and their interaction capabilities are used to achieve a balance against the game's tasks, which are adaptable parts of its content. This balance is modelled as an optimization problem, which allows for fast determination of a suitable content adaptation. The optimization problem can be solved using different approaches ranging from mathematical optimization to evolutionary algorithms depending on the given problem complexity. In order to go from a theoretical model to an applicable algorithm a set of distinct properties is chosen that suffices to model some of the most complex mechanics present in role-based multiplayer games showing the versatility of the developed adaptation model. This provides game designers using the adaptation model a toolbox for flexible content creation. This paper presents an adaptation system containing a mathematical formalization of player roles and game content as well as template modelling properties to assist in content creation. The system has been implemented in Matlab, connected to Unity and evaluated regarding different application scenarios using real data from current role-based games. [ABSTRACT FROM AUTHOR]