Real traffic data shows that there exists bottleneck 1 kilometer downstream on-ramps and traffic jams often happen at this area. Since 1960's, ITS researchers have started to pay attention to this phenomenon and study how ramps influence on freeway. Up to now, there developed many ramp control methods such as Fixed-time strategies, Reactive ramp metering strategies and Nonlinear optimal ramp metering strategies. However, these strategies hardly meet the demand of ramp control. In 1990's some observers investigated the impact of driver behavior on traffic models and provided good explanations to some problem in freeway traffic. Combining this behavior theory, this thesis first analysis the driver behavior near on-ramps and then model these behaviors at micro lever. The main work in this thesis as follows: 1. Analyzing the on-ramp traffic and studying the driver psychology Static and dynamic traffic character is proposed here and we classify the drivers on the freeway as five types by fuzzy cluster method. 2. Modeling driver behavior in three regime of traffic flow In this section we model the driver behavior in uncongested regime, congested regime and mixed regime respectively. A static model is conducted to describe the driver behavior in uncongested regime. Two-main behaviors on the freeway as" car-following and lane-changing are proposed in the behavior model of congested flow. In the end a three inputs one outputs fuzzy controller is used to probabilistically estimate the mixed regime likelihood in the other two regimes. 3. Simulating behavior models and discus the results In the simulation, the data used in our model were taken during an afternoon peak from th segment of westbound Gardiner Expressway near the Spadina Avenue on-ramp. We discus the feasibility of our behavior models by comparing the models' output flow with the actual traffic flow and also proposed the model and discus its significance in analyze the behavior on the freeway.