Abstract: The settlement of the Great Wall formed in the Ming dynasty is a typical defensive settlement in northern China, whose distribution essentially is the product of games between farming culture and nomadic culture and forms the foundations of cities and towns in the transition areas. However, the traditional research approaches restrict the deep understanding of dynamic mechanisms. Based on the multi-agent modelling system, the model of location selection for settlements was proposed from the perspective of games between farming culture and nomadic culture. The intrinsic motiration to form settlement patterns was discussed by a bottom-top generation way, taking distribution structure as the verified objecture. The simulation results showed there was a strong correlation between the micro-assistance behavior and macro-settlement distribution, i.e. self-organization of the great wall. Faced to the Great Wall, the settlement pattern in hierarchical radiation was the optimal choice in the effciency principle to balance resources against defensive requirements. The model had better global convergence and computational robustness, which was proved that agent-based simulations had potential applied values in the field of landscape patterns for settlements.