Weighted SCAN for modeling cooperative group role dynamics

Social agents have the ability of communicating and forming groups with each other. Group members in games typically share the same role. In dynamic environments with the presence of obstacles and barriers separating members from each other presents a situation where a member separated from the rest of the group, while still a member of that group, should not have the same role or updates of the rest of the group due to the physical distance presented by the obstacles. This study introduces a weighted version of the SCAN and hierarchical SCAN graph clustering algorithms which are essentially based on neighbors. The autonomous agent players in spatial strategy game scenarios tested with the weighted SCAN have demonstrated an improved realistic behaviour in the social test settings.