Feedback

The traffic simulation needs input from the demand generation, since it executes the plans from the demand generation. However, the demand generation depends on the traffic simulation because for example congestion only shows up in the traffic simulation, and demand adjusts to such shortages. In order to deal with this situation, one iterates between demand generation and traffic simulation. For example, demand generation is run assuming no congestion, the resulting traffic simulation is run, then the demand generation is run again now including the congestion from the last traffic simulation run, etc., until a steady state is reached. That is, the system is systematically relaxed towards a consistent state.

Fig. 2.4 shows an example of replanning. The traveler first changes his/her route, presumably in adaptation to congestion. Eventually, he/she decides that the destination is too far away and switches to a nearer location. Fig. 2.5 shows a systemwide consequence of replanning. The scenario is one where 50000 travelers starting at random locations all over Switzerland travel to Lugano, which is south of the Alps. The scenario is for testing purposes, but it has some resemblance with vacation traffic in Switzerland. In the initial run (left), all travelers have planned their routes assuming a completely empty network; in consequence, they all use the freeways as much as possible. After many iterations (right), travelers have learnt that because of the congestion other paths may be advantagous; as a result, traffic is much more spread out.

It should be noted at this point that there is no a priori reason why a real system should be relaxed. For example, during unique events such as trade shows or soccer games, the transportation system is probably not relaxed. The research here just follows the usual path in such situations: First understand the steady state solution, and then move on to the transients. Note that the steady state here refers to the comparison from one iteration to the next, not to a steady state across time-of-day.

Figure 2.4: Result of day-to-day learning in a test example. LEFT: Situation at 9:00am in the initial run. RIGHT: Situation at 9:00am in the 49th iteration. Each pixel on the road is a car (by overlapping in the graphics they form the traffic streams); the circle denotes where they are going. Clearly, the system has found a better solution after 49 iterations.

Figure 2.5: Feedback