To withstand the Antarctic cold on open land for more than two months, Emperor penguins are forming densely packed huddles with a hexagonal lattice structure. Video recordings have revealed striking dynamical reorganization processes within those huddles (PLoS One, 6:e20260, 2011), including wave-like patterns, global rotatory motions and abrupt transitions to a disordered state. Here we show that ba- sic aspects of the huddling behavior can be reproduced with simple systems of interacting point particles. For a more realistic modeling, the individual animals are treated as self-driven, information process- ing agents with situation-dependent behavior, similar to simulations of collective swarm behavior in fl ocks and herds. We present a multi- agent simulation in which both the spontaneous huddle formation and the observed wave patterns emerge from simple rules that only encom- pass the interaction between directly neighboring individuals. Our model shows that a collective wave can be triggered by a forward step of any individual within the dense huddle. The group velocity of the resulting wave is dependent only on the reaction times and the step velocity of the animals. By including the mutual adaption of individ- ual body orientations, we present fi rst results on rotary and curved movement patterns.