Self-Similarity in Game-Locked Aggregation

A collective game is studied via agent-based modeling approach, where a group of adaptive learning players seek for their best positions on a vertical line. The movements of players are driven by benefits obtained from interactions. The game falls into an evolutionary stable state, at which aggregations of players on the line emerge. The pattern of these aggregates exhibits self-similarity at different scales with a fractal dimension of 0.58. The underlying mechanism of this aggregation is unique in that aggregates are resulted from mutual lock-in of players. This game-locked aggregation, in contrast with the diffusion limited aggregation, is applicable to a broader scope of aggregation processes.