Using dynamic potential or dynamic partial pedestrian routes

Typical use cases of the dynamic potential function include that there is a continuum of directions to choose from. For partial pedestrian routes, the opposite applies: they are used, when there is a finite and usually small number of routing options (Route choice method for partial pedestrian routes).

The dynamic potential function could also be used to make discrete choices (e.g. whether to choose the left or right door) and partial pedestrian routes could be used to approximate continuous choices by putting areas closely next to each other via which pedestrian routes could run and by having one partial pedestrian route leading over each of these areas. However, both have disadvantages:

• In the first case, partial pedestrian routes can be used to select a small number of route choices. The computational load of pedestrian routing decisions regarding partial pedestrian routes is much smaller compared to the computational load of the dynamic potential. The simulation runs faster when route choices are simulated with partial pedestrian routes instead of dynamic potential wherever possible.
• In the second case, it is time-consuming to model a great number of partial routes just to approximate a continuous directional choice.

Exceptions from that rule:

1.  If the set of discrete route choices is not fixed, for example when zones with very high pedestrian density occur and disappear dynamically during the simulation. While a dynamic potential would automatically "react" to this, this would be difficult to capture with partial pedestrian routes.

2.  With formula-based partial pedestrian routes it is possible to consider many different influencing factors on route choice, whereas the dynamic potential function only uses the spatial distribution of pedestrians and obstacles as input. Typically such input data (e.g.signal states) are discrete anyway, yet the high flexibility of partial pedestrian routes with regard to input data may be a reason why they are the preferred option, even if other aspects would suggest to use the dynamic potential function.