Abstract
In recent years, the importance of Car-following models has increased significantly, although car-following behavior has been studied for an half of a century. In most cases, the simulation model results are used in actual field. However, correct verification about their result is not attained.
Most models do not consider about reaction time as various and in reality, it is general to use fixed value or random variable for the reaction time. As a result, it causes the rigidity of models, and the “time difference” between the actual behavior and models so, acceleration and deceleration at time (t) are simultaneously. Therefore, reaction time is used as lognormal distribution, and “perceptual threshold” was estimated from the acceleration rate versus relative speed to improve acceleration and deceleration rate.
Goodness of fit of the model is examined by Root Mean Squared error (RMSE) comparing to other models. Consequently, the new model performed better description of actual driver's behavior.
