This paper proposes a layout design method by integration of multi-agent based simulation and optimization. The proposed method adopts ‘optimization for simulation’ approach to integrate simulation and optimization, in which optimization is executed first, then the result is evaluated through simulation by various scenarios. The optimization is repeated sequentially using an output data derived by the simulation executed in each previous turn. The proposed method is applied to a layout design problem for underground shopping streets, which is the problem determines store positions to be located. The objective of the layout design is to arrange store positions attractively for passers considering their preferences and target directions, which depend on the day of the week, time zones and seasons. Computer experiments are conducted to confirm the effectiveness of the proposed method.