Statistical analyses of observed volumes are usually employed in order to estimate capacities of highway elements. Capacity of weaving sections, however, is affected by many variables such as four directions of flows and variety of lane configurations, and it is difficult to estimate capacity of weaving sections through statistical analysis based on observed capacity data for various geometries and flow conditions. It is therefore considered that the capacity value may be reproduced by the aggregation of individual driver behaviors. Using vehicle trajectory data obtained from observations on several weaving sections on urban expressways in Tokyo area, such microscopic variables as spacings and relative speeds of weaving vehicles were analyzed. Based on these analyses, a simulation model which intends to estimate weaving capacity was developed. Such maneuvers of weaving vehicles as lane changing and acceptable gap searching which affect capacity were particularly precisely modeled. The model was validated for two congested weaving sections of the Tokyo Metropolitan Expressway. Simulated values of capacities were found to fit well with the observed ones.