In this paper, we introduce the notion of a community of autonomous units as a rulebased and graph-transformational device to model processes that run interactively but independently of each other in a common environment. The main components of an autonomous unit are a set of rules, a control condition, and a goal. Every autonomous unit transforms graphs by applying its rules so that the control condition is satisfied. If the goal is reached the resulting transformation process is successful. A community contains a set of autonomous units, an initial environment specification, and an overall goal. In every transformation process of a community the autonomous units interact via their common environment. As an example, the game Ludo is modeled as a community of selfcontrolled players who interact on a common board. The emphasis of the presented approach is laid on the study of the formal semantics of a community as a whole and of each of its member units separately. In particular, a sequential as well as a parallel semantics is introduced, and communities with parallel semantics are compared with Petri nets, cellular automata, and multiagent systems.