next up previous contents
Next: The queue model for Up: More realistic CA traffic Previous: Validation of rules for   Contents

Discussion

In this chapter, we have further discussed improvements to the CA traffic simulation. It turns out that, for car traffic, such models consist of only four aspects:

Once these four aspects are implemented in a reasonable way, one has a basic model. From here on, considerable work is necessary to calibrate and validate individual details. In particular, lane changing needs to include lane changing to reach a particular lane for a turn, and lange changing on merge/acceleration lanes.

A problem with such a microsimulation approach is that the necessary input data is often not available. For example, as a minimum one needs lane connectivities (which incoming lanes are connected to which outgoing lanes, Fig. 17.8), and signal plans. Furthermore, although it is an advantage that such simulations generate link capacity instead of taking it as input data, considerable adjustments need to be done. For example, the Gotthard tunnel, as a 1-lane road without traffic light, should have a capacity of 2000 vehs/hour. According to the local police, however, the capacity not more than half of that. The reason, presumably, is that the tunnel entrance has a strong uphill slope, and acceleration of vehicles is less than normal.

Figure 17.8: Lane connectivities across intersections. This information is needed for realistic multi-lane simulations.
\includegraphics[width=0.8\hsize]{connectivities-fig.eps}


next up previous contents
Next: The queue model for Up: More realistic CA traffic Previous: Validation of rules for   Contents
2004-02-02