Introduction

As explained in Chap. 2, ``learning'' or ``adaptation'' is an extremely important part of transportation simulations packages. The idea is that if the execution of a plan differs from what people had expected, then they will change their plans to adapt to what they found. For example, if congestion lets them arrive late to work, they will leave home earlier.

We will implement this in a very straightforward way: The traffic simulation will collect link travel times, and the router will use them to generate better routes. This reflects day-to-day learning, that is, travelers revise their decisions from one day to the next. This is in contrast to within-day learning, which will be treated later.

We will also allow only 10% of the travelers to replan between any given two days, in order to avoid over-reactions of the system. Such over-reactions could otherwise for example happen if alternative A was slightly faster than another one in one iteration and as a result all travelers would switch to link A, making it extremely congested. There are other ways to deal with this problem, which will also be treated later in the class.

Fig. 13.1 gives information about the data flow through the different elements.

Figure 13.1: Data flow through the simple feedback mechanism of this chapter. Reading the network files is not drawn. The thick lines are the ones which need to be done in this Chapter.

Implementation