Urban traffic congestion is a pandemic illness affecting many cities around the world. We have developed and tested an actual urban traffic simulative model (AUTM) for predicting and avoiding traffic congestion. The model includes three key components, the map and transfer (MT) conversion, optimized spatial evolution rules, and a congestion-avoidance routing algorithm. (1) The MT conversion method is proposed to get actual urban cellular spaces, which apply optimized spatial evolution rules to simulate vehicular dynamics better. (2)AUTM is proposed for simulating traffic congestion and predicting the effect of adding overpasses and roadblocks. (3)The congestion-avoidance routing algorithm is proposed for vehicles to dynamically update their routes toward their destinations, which can achieve traffic optimization in urban simulations. Extensive experimental simulations in various actual cities are carried out. Our results in this extreme case are encouraging: the traffic congestion forecasting accuracy is more than 89%, and the variance of road density forecast is less than 0.2.