Based on the multi-agent approach and the swarm platform，a macroscopic simulation of urban transportation system is developed. In the simulation，travelers are divided into three groups first，and then links，intersections and travelers are treated as agents respectively. Three different experiments，namely traveling scales，traveler groups and network structure experiments，are carried out by the simulation. Through comparing and analyzing the simulation results from the perspectives of system scale and structure，we try to find and explain the macroscopic phenomena and the possible emergence phenomena coming from the microscopic level in urban transportation system.