Sudden changes on road networks, including new roads, bridge construction, road blockage or traffic accidents cause travelers to switch their routes to less costly ones as compared to alternative routes. Travelers, however, tend to take higher cost routes due to insufficient information and errors in perceived travel time. This may cause severe congestion on a certain route. Conventional models, however, are unable to adequately simulate travelers’ behavior under such suddenly changing network conditions. The objective of this paper is to analyze travelers’ daily travel behavior in such cases via a stochastic process, the Markov-chain approach, which is considered to be a suitable method for representing sudden changes in states. This model is based on agent and we assumes that travelers select their route via learning process of travel time that they had previously experienced.