Introduction

In Chap. 7 we have introduced a simple cellular automata micro-simulation. The reason to chose that particular modelling technique was that it is conceptually simple, relatively easy to implement, somewhat realistic, and it fulfilled the functionality that was needed at that point in the project. In this chapter, an alternative will be presented, the so-called queue model (46). For experts: The queue model is essentially a standard queueing model, but with storage constraints added. Storage constraints mean that links can be full, which causes spillback across intersections.

The queue model is in our view the simplest dynamic model that is somewhat useful for real world predictions (see Chap. ). Despite some obvious shortcomings in the description of the dynamics (see Chap. ) in particular with respect to traffic jam wave backpropagation, we are not aware of any empirical evidence showing that more sophisticated models are truly better with respect to their predictive power. However, the path to more realistic simulations does not go via the queue model, but is a continuation of the explicit spatial methods, such as the CA. Making those methods, possibly on continuous rather than cellular space, useful for the real world (Chap. 17) is considerably more work than making the queue model useful for the real world. In consequence, if one intends to use the methods presented in this text for real world applications, one needs to carefully weigh advantages and disadvantages: The queue model of this Chapter is the fastest path to some usefulness, but is eventually limited; the CA model of Chaps. 7 and 17 (or non-cell based variants of this) are considerably more work but ultimately more realistic and more flexible.