Abstract: In order to get better description of real traffic, an improved cellular automaton (CA) model is proposed on the basis of a sensitive driving CA model. The slow-to-start effect is introduced during the acceleration process, which is dependent on acceleration probability. It is found that the maximum flow decreases with decreasing acceleration probability. Furthermore, a refined sensitivedriving model is proposed. The corresponding simulation indicates that, with the density lower than 0.24, the flow is independent of acceleration probability, while with the density higher than 0.24, the flow reduces with decreasing acceleration probability for the same density. Phase separation between free phase and congested phase is clearly worked out in a temporalspatial pattern. Moreover,the meta-stability phenomena are reproduced under different initial conditions.Finally, the simulated fundamental diagram shows a pretty good agreement with that based on experimental data.


Key words: acceleration probability, fundamental diagram, sensitive driving, slow-to-tart, cellular automaton (CA)