With rapid development of China's economy and urbanization, the fast growth of private car has led to excessive burden on urban roads. As an important way to control traffic congestion, traffic demand management has to be based on traffic demand prediction. This paper establishes a set of activity-travel demand models based on multi-agent simulation which is suitable for the characteristics of Chinese travelers. Based on the historical data of people's travel in Jiangning District, Nanjing, training set and test set are built, which are used for model parameter estimation and case study respectively. By importing the travel demand data obtained from the model into the open source software MATSim for traffic allocation, the complete process of activity-based model is realized. By comparing the prediction results with the measured travel data, we can find that the activity-based model can achieve higher accuracy in travel demand prediction.
 

Travel demand prediction;Activity-based model;Multi-agent simulation;Traffic allocation